Book Preface Atlas of Human Anatomy uses Frank H. Netter, MD’s detailed illustrations to demystify this often intimidating subject, providing a coherent, lasting visual vocabulary for understanding anatomy and how it applies to medicine.
This 5th Edition features a stronger clinical focus-with new diagnostic imaging examples-making it easier to correlate anatomy with practice. Student Consult online access includes supplementary learning resources, from additional illustrations to an anatomy dissection guide and more. It’s how you know. New to This Edition. Expand your study at studentconsult.com, where you’ll find a suite of learning aids including selected Netter illustrations, additional clinically-focused illustrations and radiologic images, videos from Netter’s 3D Interactive Anatomy, dissection modules, an anatomy dissection guide, multiple-choice review questions, “drag-and-drop” exercises, clinical pearls, clinical cases, survival guides, surgical procedures, and more. Correlate anatomy with practice through an increased clinical focus, many new diagnostic imaging examples, and bonus clinical illustrations and guides online.
For decades rodents have been used to explore normal brain functions and mechanisms underlying brain diseases. Such data often have been the basis in the search for new drugs. In this study we selected chemical markers associated with central noradrenaline and serotonin neurons, key systems in research on and current treatment of depression, and studied their expression with in situ hybridization in postmortem human brains. The results show some distinct species differences between human and rodent noradrenergic and serotonergic neurons which may better inform the development of novel anxiolytic/antidepressant drugs. Using riboprobe in situ hybridization, we studied the localization of the transcripts for the neuropeptide galanin and its receptors (GalR1–R3), tryptophan hydroxylase 2, tyrosine hydroxylase, and nitric oxide synthase as well as the three vesicular glutamate transporters (VGLUT 1–3) in the locus coeruleus (LC) and the dorsal raphe nucleus (DRN) regions of postmortem human brains. Quantitative real-time PCR (qPCR) was used also. Galanin and GalR3 mRNA were found in many noradrenergic LC neurons, and GalR3 overlapped with serotonin neurons in the DRN.
The qPCR analysis at the LC level ranked the transcripts in the following order in the LC: galanin GalR3 GalR1 GalR2; in the DRN the ranking was galanin GalR3 GalR1 = GalR2. In forebrain regions the ranking was GalR1 galanin GalR2. VGLUT1 and -2 were strongly expressed in the pontine nuclei but could not be detected in LC or serotonin neurons. VGLUT2 transcripts were found in very small, nonpigmented cells in the LC and in the lateral and dorsal aspects of the periaqueductal central gray. Nitric oxide synthase was not detected in serotonin neurons.
These findings show distinct differences between the human brain and rodents, especially rat, in the distribution of the galanin system and some other transmitter systems. For example, GalR3 seems to be the important galanin receptor in both the human LC and DRN versus GalR1 and -2 in the rodent brain. Such knowledge may be important when considering therapeutic principles and drug development.
The LC/subcoeruleus is a bilateral compact accumulation of neurons in the pons with a limited rostro-caudal extension and a characteristic localization close to the fourth ventricle. This nucleus is characterized by the expression of the catecholamine-synthesizing enzyme TH, here visualized through the TH transcript ( and ). In the dark-field configuration the NA neurons can be distinguished by their autofluorescence. Therefore, in several instances we used the pigment as a marker for NA-LC neurons instead of using TH ISH.
In contrast, the 5-HT neurons expressing the TPH2 transcript in the caudal mesencephalon/pons ( and ) extend over a long distance with characteristic distribution patterns at different levels. This group is localized in the midline ventral to the aqueduct, extending laterally in two wings in the ventral PAG (vPAG) and continuing further ventrally as MRN with many neurons dispersed in the pontine reticular formation. Dark-field ISH photomicrographs showing the distribution of galanin, GalR1, GalR3, and nNOS in the LC. Autofluorescent pigmented neurons are indicated by arrowheads, and positive cells are indicated by arrows. The three markers TH ( A and B), galanin ( C and D), and GalR3 ( E and F) show an overlapping distribution pattern, whereby TH and GalR3 transcript levels are similar in all cells. Note the variability in the strength of the signal for galanin mRNA.
Insets A′, C′, and E′ show results after hybridization with sense/control probe. Strong signal for GalR3 is seen in LC ( G) but not after hybridization with the control probe ( H). Note bluish autofluorescence from pigment in NA-LC neurons (arrowheads in H). ( I) A very weak signal for the GalR1 transcript is seen in the LC, but it is probably not present in the NA neurons. NNOS mRNA overlap with TH + cells in the LC only to a minor extent in its most rostral part ( J), where pigmented neurons intermingle with NOS + neurons ( K). Exposure time: TH, 10 d; galanin, 4 wk; GalR1 and GalR3, 8 wk; nNOS, 6 wk.
(Scale bars: 200 μM in A, C, E, and J; 100 μM in B, D, and F; 50 μM in I and H; 25 μM in G and K.). Dark-field ISH photomicrographs showing the distribution of TPH2 ( A and D), galanin ( B and E), and GalR3 ( C and F) in semiadjacent sections. D, E, and F show higher magnification of boxes in A, B, and C, respectively. Arrows indicate cells positive for the respective marker. Note overlapping distribution of TPH2 and GalR3 in the posterior DRN (compare D and F).
However, no galanin signal is seen in that particular area ( B and E). Instead galanin+ cells are seen more dorsally ( B). Exposure time: TPH2, 10 d; galanin, 4 wk; GalR3, 8 wk. (Scale bars: 100 μM in A, B, and C; 50 μM in D, E, and F.) The NA-LC neurons express robust levels of TH, which can be observed after comparatively short exposure times and are overexposed after only 10 d, as compared with the less strong expression of galanin) (4-wk exposure time) ( and the weakly expressed GalR3 (8-wk exposure time). Thus, it is the varying exposure times that make the signals in look approximately similarly strong. Note that GalR3 transcript levels are fairly similar among individual neurons , as is the TH signal , in contrast to the high variability of galanin mRNA. GalR3 appears to be the predominant galanin receptor in the LC.
A weak GalR1 signal was seen in the LC region expanding medially into the pontine central gray but did not seem to be associated with the pigment characteristic of the NA neurons as confirmed in cresyl violet-counterstained sections analyzed with bright- and dark-field microscopy (see ). A comparison of the distribution of transcripts for TPH2 , galanin , and GalR3 in the DRN/vPAG shows that TPH2 and GalR3 exhibit a distinct overlap, but no galanin mRNA can be seen in this particular subregion. (Exposure times for TH, galanin, and GalR3 were approximately as stated above.) However, galanin mRNA is observed at this level just dorsal to the serotonin cells and extending dorsally/laterally in the vPAG. Galanin mRNA was also seen in a number of additional nuclei, including the inferior colliculus , the tegmental peduncular pontis , the medial parabrachial nucleus , and the subcuneiform nucleus.
GalR1 and -3 were also observed in other regions in the sections analyzed; e.g., a robust GalR1 signal was observed in the lateral PAG , and GalR3 signal was seen lateral to the aqueduct. A signal for GalR2 mRNA could not be detected in any of the sections.
NNOS was expressed in many regions of the brainstem, but these cells overlapped with TH-positive (+) mRNA in the LC only to a minor extent and only in its most rostral part, where pigmented NA neurons intermingled with NOS + neurons. There was no overlap with the TPH2 + 5-HT neurons in the DRN or MRN (compare A and B in ).
However, strongly labeled large neurons were seen in the ventro-lateral PAG extending into the reticular formation, and many small, weakly labeled cells were encountered in the dorso-lateral region of the PAG. Dark-field ISH photomicrographs showing the distribution of nNOS ( A) and TPH2 ( B) transcripts in semiadjacent sections. There is no overlap with the TPH2 + 5-HT neurons in the DRN. Aq, Aqueduct; MLF, medial longitudinal fascicle. Exposure time: TPH2, 10 d; nNOS, 6 wk.
(Scale bars: 100 μM.) VGLUT1 mRNA could not be detected in the vPAG, including the DRN. However, a very strong VGLUT1 mRNA signal was observed in many neurons of the pontine nuclei. VGLUT2 also was strongly and frequently expressed in these nuclei, overlapping with VGLUT1 mRNA. At the level of LC, VGLUT2 mRNA was seen in the pontine central gray substance, in the LC area, and extending medially into the region of the dorsal tegmental nucleus.
High-power magnification showed that in most instances grains did not appear to overlie the autofluorescent noradrenergic cell bodies but instead represented a separate population of very small cells, possibly glial cells. In contrast to VGLUT1, there was a strong VGLUT2 signal in the PAG, primarily in the lateral/dorsolateral region. The midline area lacked detectable transcript, in strong contrast to the distribution of TPH2 mRNA (compare F and G in ). Dark-field ISH photomicrographs showing the distribution of VGLUT1 ( A– C), VGLUT2 ( D, E, G, and H), and TPH2 ( F) mRNA at the pontine level. The box in F shows approximately the region displayed in the semiadjacent section in G. There is no detectable signal for VGLUT1 in the vPAG, including the DRN ( A), whereas there is a very strong signal for both VGLUT1 ( B) and -2 ( C) in the pontine nuclei.
Online Atlas Of Anatomy
A distinct VGLUT2 signal is seen in the pontine central gray, including the LC, extending medially to the dorsal tegmental nucleus ( D). As seen in the high-power magnification, the cells are very small (arrows) compared with the pigmented NA neurons (arrowheads in E). Numerous vGLUT2 mRNA+ cells are observed in the lateral PAG ( G and H), although there are very few cells in the DRN and vPAG (compare G and F). MLF, medial longitudinal fascicle; PN, pontine nuclei. Exposure time: TPH2, 10 d; VGLUT1 and VGLUT2, 8 wk. (Scale bars: 200 μM in A, D, and F; 100 μM in B and G; 50 μM in H; 25 μM in C and E). The relative expression of galanin and its receptors was analyzed by qPCR in tissue slices of containing LC and DRN and in macrodissected tissue punches from frontal cortex, cingulate cortex, and the amygdaloid complex (for details, see ).
In addition, TH raw cycle threshold (Ct) value 28 and TPH2 (raw Ct value 25) mRNA levels also were analyzed in such slices. Galanin mRNA expression was ∼42-fold higher than GalR1 in the LC slices and was ∼25-fold higher than GalR1 in the DRN slices. GalR3 was the most abundantly expressed galanin receptor, being approximately sixfold higher than GalR1 in LC slices and approximately sevenfold higher in DRN slices. GalR2 was expressed at very low levels. The raw Ct values for galanin and GalR1-3 transcripts are given in. QPCR was used to examine the mRNA levels of galanin, GalR1, GalR2, and GalR3 in the LC ( A) and DRN ( B).
The expression level for each gene examined was normalized to β-actin and then expressed relative to GalR1, the value for which was set at 1. Expression levels for galanin and GalR3 in the LC and DRN are significantly higher than those of GalR1 and GalR2, which are very low. The bar graphs represent the mean ± SEM ( n = 7). Statistical significance was determined using unpaired Student’s t test. P 35 for both. Taken together, these results indicate that GalR3 seems to be the abundantly expressed receptor in the brainstem nuclei studied here, whereas GalR1 appears to predominate in the forebrain regions.
Discussion Almost 30 y ago Tatemoto, et al. discovered galanin, and subsequently three receptors were identified, GalR1-3 (, ), which are the focus of the present study. Here we studied these molecules in selected areas of human postmortem brains with ISH and qPCR to establish expression patterns and to reveal possible differences among species. Some other transmitter-related markers, i.e., TH, TPH2, nNOS, and VGLUTs, were included also.
Subjective rating ranging from weak (+), medium strong (), strong (), to very strong () signal. N.d., not detected. Data are from the present study and papers cited in the text. It should be noted that we report only transcripts without evidence for the expression of the receptor protein. Thus, GalR2 and -3 receptors may be present on afferents (presynaptic) to the forebrain regions studied. For example, there is evidence that in the rat the NA-LC neurons synthesize GalR2, which probably is transported centrifugally to act as a presynaptic receptor in forebrain regions.
Galanin System: LC. The present results confirm that many neurons in the LC in the human brain robustly express galanin and TH (for references, see the Introduction). The analysis was facilitated by the fact that the NA-LC neurons express a distinct pigment and thus do not need to be identified by TH staining. Importantly, a strong GalR3 signal but no GalR1 signal was detected in these neurons. However, a weak GalR1 mRNA signal was seen in non-NA cells in the LC and in some other brain regions, apparently, the GalR2 transcript could not be detected. In agreement, the qPCR results confirmed that the GalR3 transcript was higher than GalR1 and GalR2 at the LC level.
Galanin itself was robustly expressed. The marked variation in galanin mRNA levels among individual LC neurons, in contrast to the apparently similar levels of GalR3 (and TH), probably reflects the fact that galanin is a released molecule that must be replaced through compensatory synthesis, i.e., requires new transcript. This variation also may indicate that individual LC neurons have different activities/functions. GalR3, on the other hand, like the enzyme TH, is a comparably stable protein, a receptor with a low turnover rate. The GalR3 result was surprising because, based on ISH, GalR1 and -2, but apparently not GalR3, are present in the rat and mouse LC (, –). In fact, GalR3, the last cloned receptor (, ), has a comparatively limited expression in the rat brain (, ).
The Allen Brain Atlas provides no information about GalR3 in the mouse brain , possibly reflecting absence or low levels. Whether our riboprobe(s) that were selective for GalR2 failed or whether GalR2 indeed is not present in the human LC remains to be clarified. In the qPCR analysis of LC, GalR2 had the highest Ct value, suggesting a very low, if any, expression (see below). Galanin System: vPAG/DRN. Some galanin cells with a fairly weak signal were seen in the midline and in the vPAG but extended mainly dorsolaterally and did not overlap with the 5-HT neurons. Numerous galanin+ neurons also were seen in the same sections, e.g., in the inferior colliculus and other regions. These results and the LC data suggest that the galanin probe was functional.
In contrast, there was a strong overlap between 5-HT and GalR3 mRNA + neurons, indicating coexistence. However, we failed in attempts to carry out double-ISH to provide final evidence for coexistence. One explanation for the failure may be that the exposure time to obtain a robust GalR3 signal was more than 2 mo. Taken together, the results indicate that in humans galanin is expressed strongly in NA-LC neurons but not in 5HT-DRN neurons, similar to expression in the mouse but in contrast to expression in the rat. The LC and probably the DRN express GalR3 in humans, thus representing targets for GalR3 antagonists/agonists. The qPCR data agree with the ISH results, with galanin being the most abundantly expressed molecule, followed by GalR3 and then GalR1 and -2. Galanin System: Forebrain.
A limited qPCR analysis of the transcripts for the galanin systems was carried out in two cortical regions and the amygdala. Interestingly, the levels of all markers were generally low raw Ct values ranged from ∼40 (GalR3) to ∼31 (GalR1); the GalR1 transcript showed the highest levels, even higher than galanin itself.
Thus, the early 125I-galanin binding studies on some monkey and human forebrain areas (–) might have visualized mainly GalR1 receptors, although receptors on afferents represent a possibility (see above). These results indicate an interesting situation: GalR3 is associated with projection neurons in the lower brainstem but may be of little importance in the forebrain, where GalR1 seems to predominate. It is likely that GalR1 and perhaps GalR3 are postsynaptic receptors. Differences Among Species in Other Galanin Systems. Differences among species have been noted for other galanin systems.
In the rat the cholinergic forebrain neurons, of interest in relation to Alzheimer’s disease , express galanin after the inhibition of axonal transport with colchicine (, ) and after brain injury (–), although the galanin transcript can barely be detected in these neurons in the normal rat. Colchicine also is needed to detect galanin peptide in these neurons in mouse. However, strong expression can be seen in some monkeys e.g., owl monkeys , capuchin monkeys and rhesus monkeys but not in humans. Thus, with regard to this system, the rat seems to be more similar to humans than to the lower monkeys. To what extent human cholinergic forebrain neurons have the capacity to express galanin, as seen in the rat and the mouse, is still unclear. As a note, Chan-Palay et al.
reported the presence of galanin mRNA + neurons in the human forebrain, but these neurons probably were not identical to the cholinergic neurons. Nevertheless, Mufson and colleagues have reported impressive data providing evidence that galanin has a protective role in Alzheimer’s disease.
A similar situation may exist in the histaminergic/GABAergic tubero-mammilary neurons that express galanin after colchicine administration in the rat (, –) but apparently not in human, although this analysis was performed only with immunohistochemistry and not with ISH. NNOS and Glutamate. Neither nNOS nor VGLUT3 could be detected in human 5-HT neurons. The lack of nNOS in human is similar to mouse (, ) and different from rat (–, ) but is in agreement with a study by Carrive and Morgan based on NADPH-d histochemistry showing strong staining in the dorsal PAG but apparently none in the DRN. NAPDH-d in the rat has been shown to be identical to nNOS (, ).
Here we provide confirmatory evidence by showing the apparent absence of nNOS transcript in human 5-HT neurons based on ISH but a strong signal in the nearby regions, indicating a working probe. Also the NA-LC neurons had no detectable nNOS signal, even if nNOS + cells intermingled with NA neurons to a limited extent. The discovery of VGLUTs (–) made it possible, via immunohistochemistry and ISH, to identify unequivocally the neurons using the excitatory amino acid glutamate as transmitter. However, even though powerful antibodies to the VGLUTs have been generated, VGLUT expression in neuronal cell soma can be visualized only with ISH, because antibodies apparently show the transporters only in nerve endings. DRG neuron cell bodies are an exception (, ). The distribution patterns in mouse and rat DRN/PAG are similar (for references, see Introduction): no VGLUT1 mRNA is found in the LC, DRN, or vPAG.
There are many VGLUT2 mRNA + cell bodies in the rat and mouse vPAG but not in the LC or in the midline of the vPAG, i.e., the DRN. In contrast, a distinct subpopulation of 5-HT neurons in the DRN express VGLUT3 transcript in rats (–), mice , and syrian hamsters. However, not all rat VGLUT3 + neurons in the DRN are serotonergic. In humans, as in rats and mice, no VGLUT1 transcript was found in the LC region or in the vPAG.
VGLUT2 mRNA was found in small cells (possibly glia) in the LC, partly intermingling with NA neurons, and was seen only in the dorsal and lateral PAG, in contrast to the distribution throughout the entire PAG in rodents. The lack of a VGLUT3 signal may represent either a false negative caused by an unsensitive or failed probe or a true absence. Thus, it still is uncertain whether VGLUT3 expression in the 5-HT neurons in the vPAG differs in rodents and humans. Comparison with Results from the Allen Institute for Brain Research. Scientists at the Allen Institute for Brain Research also have explored the human brain (, ), as described in. The data analysis revealed some individual variability, but, in general, galanin expression was higher in the supraoptic nucleus, hypothalamus, LC, and frontal lobe than in the central nucleus of the amygdala and MRN.
TH and TPH2 transcripts also were highly expressed in LC and MRN, respectively. Expression levels for the galanin receptors were generally low and are difficult to compare with the present qPCR and ISH results. Taken together, the results from the Allen Institute are, to a certain extent, in agreement with the qPCR and ISH results in our study. However, this comparison needs further, in-depth analysis.
Nonisotopic ISH for galanin and GalR2 also has been performed in cortical and subcortical regions, where a weak signal is observed for both transcripts. Methodological Aspects. Histochemical approaches always are associated with methodological problems related to sensitivity and specificity (, ). These difficulties are accentuated when studying human postmortem tissue because of variability among human beings, deterioration of the tissue with postmortem time, varying conditions of death, and subsequent handling. Even more problematic are low-abundance transcripts, e.g., for certain receptors. Such proteins have slow turnover and thus low mRNA levels because they, unlike neuropeptides, do not need to be replaced swiftly after release. Presumably such receptors can be revealed only by advanced molecular biology combined with electrophysiology (, ).
In the present study, the transcripts for robustly expressed molecules, such as TH and TPH2, could be demonstrated in all brains but, importantly, after considerable differences in exposure time to an autoradiographic film/emulsion. Thus, although sections from some brains hybridized for TH or TPH2 needed less than a week of exposure, sections from other brains had to be exposed for several weeks. Not unexpectedly, such differences were critical for the successful processing of the low-abundance galanin receptor transcripts.
Thus, in principle, such transcripts could be detected only in brains with short exposure time for TH and TPH2. Therefore, negative results should be interpreted with caution. The probe also is essential. We obtained good results for GalR3 with only one of the designed probes.
The probe may not be as good for GalR1 for GalR3, and the probe may have failed completely for GalR2 and VGLUT3. Again, qPCR results indicate very low GalR1 levels. As a note, GalR2 and -3 have a high degree of homology; therefore in a previous study we demonstrated the specificity of the primers for the two receptors. Difficulties in detecting galanin receptor transcripts are supported by the recently published Allen Brain Atlas describing the distribution of some 20,000 transcripts throughout the mouse brain : No results are reported for GalR3, GalR2 is distinctly expressed only in some of the total of 52 sections/levels, and GalR1 appears less abundant than in rat brain.
Download aplikasi penjualan tiket berbasis web. Aplikasi ini adalah aplikasi pemesanan tiket bus pada buku Bikin Aplikasi Pemesanan Tiket Online Berbasis Web dan Android terbitan CV. Jadwal bus demo: Semarang Jogjakarta 2015-08-15 Semarang Jogjakarta 2015-08-15 Semarang Wonosobo-Purwokerto 2015-02-08 Semarang Wonosobo-Purwokerto 2015-02-08 Semarang Wonosobo-Purwokerto 2015-02-08 Semarang Wonosobo-Purwokerto 2015-02-08 Semarang Tegal 2015-08-30 Semarang Tegal 2015-08-30 Semarang Tegal 2015-09-28 Semarang Jogjakarta 2015-09-28 Semarang Jogjakarta 2015-09-28. Asfa Solution.
However, the presence of the GalR3 transcript has been reported in many rat brain regions by Mennicken et al. using ISH and by Waters and Krause using qPCR. In contrast, the Allen Brain Atlas shows robust expression of many other neuropeptide receptors and thus supports our view that galanin receptor transcripts may be particularly difficult to visualize with ISH and, as shown here, probably are even more difficult to visualize in human postmortem tissue. The qPCR analysis was carried out on full coronal sections of the pons, from a level including either the DRN or LC; in fact, these sections were alternate sections to those taken for ISH. However other regions within the slice expressing the studied markers to a varying extent have been included also.
For the LC level, TH probably is confined to the LC and subcoeruleus; GalR1 and -3 were observed only in relation to the LC, whereas galanin was expressed in several other nuclei. TPH2 is fairly widely expressed both in the DRN and MRN and in the reticular formation.
Galanin and GalR1 also are expressed outside the raphe region, but we observed GalR3 only in the DRN and lateral to the aqueduct. Thus TH mRNA is diluted, and results for galanin and GalR1 are not selective for the LC region. In fact, a preliminary qPCR experiment with laser capture microdissection (LCM) of the NA-LC nucleus (defined by pigmented neurons) shows a higher level of TH transcript expression (i.e., a raw Ct value of ∼22 for LCM versus ∼28 for the whole section). Similarly for GalR3 the raw LCM Ct value was ∼29, versus ∼33 for the whole section, thus showing the expected dilution effect for these markers. Autoinhibition, Dendro-Somatic Release, and Electrophysiology. Autoinhibition of LC neurons mediated by NA (–) and of DRN neurons mediated by 5-HT may involve not only release from collaterals but also somato-dendritic release (, ). This type of release also has been shown for neuropeptides and possibly for galanin in the LC.
It is assumed that this autoinhibitionis responsible, at least in part, for the delayed onset of the clinical effect of monoamine-reuptake inhibitors (, ). Electrophysiological studies in the rat have revealed that galanin hyperpolarizes NA-LC neurons (, ), presumably mediated via GalR1 , a transcript known to be present in these neurons (, ). In addition, galanin at low concentrations (10 −9M) enhances the autoinhibitory effect exerted by NA on LC neurons via α2A adrenoreceptors. Moreover, GalR1, but not -2 or -3, is regulated by galanin signaling in the rat LC (, ). Galanin also has an inhibitory effect on some 5-HT neurons in the DRN. However, it still is unclear if these effects are direct, via GalR1 or GalR3 (, ), indirect via GABA neurons , or both. ISH results indicate that GalR1 indeed is present in the vPAG but not in 5-HT neurons (, ).
Moreover, low galanin concentrations enhance the autoinhibitory effect of 5-HT via the 5-HT1A receptor in the DRN. This effect may be related to the formation of GalR1-5-HT1A receptor heteromers as recently reported by Borroto-Escuela et al.
Such receptor complexes increase trafficking of 5-HT1A receptors to the plasma membrane and have been suggested to contribute to development of depression. Neuropeptides and Stress-Related Disorders.
Several antidepressants exert their effect via monoamine neurons (, –). It has been proposed, based on animal experiments over the last few decades, that neuropeptide also receptors are putative targets for development of antidepressants, including receptors for substance P, neuropeptide tyrosine, corticotropin-releasing factor/corticotropin-releasing hormone, melanocyte-concentrating hormone, vasopressin, and dynorphin (–), as well as the galanin system. Thus, galanin interacts with 5-HT1A receptors, and galanin antagonists and agonists have anxiolytic and antidepressive effects in animal experiments (–). Interestingly, association of genes encoding galanin and/or GalR3 has been reported for psychiatric phenotypes (–), including panic disorder , depression-related parameters (–), and nicotine dependence (, ). The LC plays an important role in the development of stress-related disorders (–), and stress up-regulates galanin expression in the rat LC (, ). We therefore hypothesize that such disorders are associated with an increase in firing and with increased galanin synthesis and release from LC neurons, especially from soma and dendrites (, ), resulting in activation of GalR1 autoreceptors, inhibition of firing, and decreased NA release in the forebrain.
Together, these events presumably result in a prodepressive effect. In theory, a similar scenario may be true for the 5-HT/galanin neurons in the DRN. Consequently, attenuating the inhibition and autoinhibition of NA and 5-HT neurons by galanin antagonists may have anxiolytic/antidepressive effects. In fact, these effects may be enhanced by the GalR1-5-HT1A heterodimerization described above. Finally and interestingly, Murck, et al. have shown that the effects of i.v.-administered galanin on sleep EEGs in healthy subjects are similar to those seen with sleep deprivation and that galanin, when given to patients with depression, has an acute antidepressive effect. The site of action of peripherally administered galanin remains to be analyzed.
Based on the rat experiments, a GalR1 antagonist would be suitable to obtain such an effect in the LC (, ), whereas a GalR3 antagonist (, ) may be more appropriate in the DRN. Our study shows that in the human brain the NA-LC, and presumably the 5-HT-DRN neurons, express GalR3, not GalR1 as in the rat (, ). The transduction mechanism for GalR3 has not been well characterized. used Xenopus oocytes and coexpressed GalR3 with the potassium channel subunits GIRK1 and -4.
They found that galanin opens potassium channels and thus hyperpolarize the cell membrane. If the result from this artificial system should turn out to hold true for human NA LC and 5-HT DRN neurons also, then a GalR3 antagonist could have the same effect in humans as in the rat.
The investigation of galanin has been hampered by the lack of selective and powerful pharmacological tools to analyze galanin functionality, particularly drugs that can penetrate the blood–brain barrier. However, a small number of such compounds active at the GalR3 receptor, with anxiolytic and antidepressant activity in various rat models, have been developed (, ). Concluding Remarks. A main message of the present study is that there are distinct differences among species with regard to certain transmitter systems and that results in rodent models cannot always be translated directly to humans. Such species differences may be particularly common when studying the coexistence of various transmitters, particularly those related to neuropeptidergic systems , although here we show that species differences also may affect NOergic and perhaps glutamatergic systems.
However, in general, peptidergic systems are highly conserved. For example, in a comprehensive ISH study, Krolewski, et al. reported that several important neuropeptides are distributed similarly in the human and rodent hypothalamus. Nevertheless, species variations should be taken into account when developing drugs for human disorders. The present evidence for GalR3 signaling in NA-LC and, presumably, in 5-HT-DRN neurons indicates that a similar mechanism may operate in both types of neurons in humans and that GalR3 is a relevant target for drugs aiming to treat humans suffering from anxiety and/or depression. However, to our knowledge, no clinical trials have been carried out with the GalR3 antagonists mentioned above.
RNA Probe Synthesis. RNA probes specific to TH, galanin, GalR1, and GalR3 were prepared from human dorsal root ganglion (DRG) mRNA (Clontech). TPH2, GalR2, VGlut 1, VGlut2, VGlut3, and nNOS were generated from human total-brain RNA (Ambion). The human DRG mRNA and total-brain RNA were reverse transcribed to generate cDNA using the Retroscript Kit (Ambion).
Knut Miller Atlas Of Anatomy
This cDNA then was amplified using specific primers , subcloned into a PCR1II-TOPO vector (Invitrogen), and confirmed by nucleotide sequencing (KIGene). The plasmids were linearized and then transcribed using T7 and SP6 RNA polymerases to generate sense and antisense RNA probes. In vitro transcription was carried out using the MAXIscript SP6/T7 kit (Ambion) and α 35-UTP (Perkin-Elmer) according to the manufacturer’s instructions. The transcripts then were purified using NucAway Spin Columns (Ambion). Sense probes were used as negative controls. Microscopic Analysis.
Sections were analyzed using a Nikon Eclipse E600 microscope equipped with a bright- and dark-field condenser and epi-polarization with side entrance illumination (Fiberoptic-Heim AG) and epi-fluorescence with appropriate filters combinations connected to a digital camera (Nikon DXM 1200). In some cases, Kodak T-MAX 400 black-and-white film was used for photography. Sections were scanned using a Nikon LS-2000 film scanner (Nikon). Scanned and digital images were imported into Adobe PhotoShop 6.0 (Adobe Systems, Inc.) and optimized for brightness, contrast, and sharpness. The atlases of Paxinos and Xu-Feng and Olszewski and Baxter were consulted throughout this work. We thank Professor Sandra Ceccatelli, Dr.
Christina Bark and Dr. Roshan Tofighi for advice with regard to qPCR analysis and Professor Nenad Bogdanovic for valuable advice concerning the neuroanatomy of the human brain stem and Dr. Csaba Adori, Blanca-Silva Lopez, Yu Qian, and Mingdong Zhang for their valuable assistance. This study was supported by the Swedish Research Council (04X-2887); the Marianne and Marcus Wallenberg Foundation; the Knut and Alice Wallenberg Foundation; Grant NEWMOOD; LHSM-CT-2003-503474 from the European Union; the National Alliance for Research on Schizophrenia and Depression; a grant from AFA (the Swedish Insurance company); funds from Karolinska Institutet; and the Swedish Brain Foundation.
The studies would not have been possible without the earlier support of an Unrestricted Bristol-Myers-Squibb Neuroscience grant.
General Collections Keywords: ATLAS; human anatomy; Florence Fenwick Miller Credit line This file comes from, a website operated by Wellcome Trust, a global charitable foundation based in the United Kingdom. This tag does not indicate the copyright status of the attached work. A normal is still required. See for more information. References.
Photo number: L0027705. Full Bibliographic Record: Source/Photographer. Gallery:. Wellcome Collection gallery (2018-03-29): Other versions Licensing This file is licensed under the license. You are free:. to share – to copy, distribute and transmit the work. to remix – to adapt the work Under the following conditions:.
attribution – You must attribute the work in the manner specified by the author or licensor (but not in any way that suggests that they endorse you or your use of the work). CC BY 4.0 Creative Commons Attribution 4.0 true true File history. This file contains additional information such as Exif metadata which may have been added by the digital camera, scanner, or software program used to create or digitize it. If the file has been modified from its original state, some details such as the timestamp may not fully reflect those of the original file. The timestamp is only as accurate as the clock in the camera, and it may be completely wrong. Author Wellcome Library, London Copyright holder Copyrighted work available under Creative Commons Attribution only licence CC BY 4.0 Image title L0027705 Fenwick Miller, An atlas of anatomy.
Credit: Wellcome Library, London. Wellcome Images [email protected] Title page. An atlas of anatomy or pictures of the human body.
Florence Fenwick Miller Published: 1879 Copyrighted work available under Creative Commons Attribution only licence CC BY 4.0 IIM version 2 Short title L0027705 Fenwick Miller, An atlas of anatomy. Headline L0027705 Fenwick Miller, An atlas of anatomy.
An atlas of anatomy. Anatomy. The House of Life (1878). An Atlas of Anatomy (1879). Animal Physiology for Elementary Schools (1882).
Social reform. Readings in Social Economy (1883).
On the Programme of the Women's Franchise League, An Address Delivered at the National Liberal Club, Feb. 25, 1890 (1890).
Biography. The Lessons of a Life: Harriet Martineau.
A Lecture, Etc (1877). Harriet Martineau (1884). Fiction.
Lynton Abbott's Children (1879, 3 volumes) Further reading. Van Arsdel, Rosemary T., (2001), Florence Fenwick Miller: Victorian Feminist, Journalist and Educator, Ashgate Publishing Limited. References.
The immense increase in the generation of genomic scale data poses an unmet analytical challenge, due to a lack of established methodology with the required flexibility and power. We propose a first principled approach to statistical analysis of sequence-level genomic information. We provide a growing collection of generic biological investigations that query pairwise relations between tracks, represented as mathematical objects, along the genome.
The Genomic HyperBrowser implements the approach and is available. The combination of high-throughput molecular techniques and deep DNA sequencing is now generating detailed genome-wide information at an unprecedented scale. As complete human genomic information at the detail of the ENCODE project is being made available for the full genome, it is becoming possible to query relations between many organizational and informational elements embedded in the DNA code. These elements can often best be understood as acting in concert in a complex genomic setting, and research into functional information typically involves integrational aspects.
The knowledge that may be derived from such analyses is, however, presently only harvested to a small degree. As is typical in the early phase of a new field, research is performed using a multitude of techniques and assumptions, without adhering to any established principled approaches. This makes it more difficult to compare, reproduce and realize the full implications of the various findings. The available toolbox for generic genome scale annotation comparison is presently relatively small.
Among the more prominent tools are those embedded within the genome browsers, or associated with them, such as Galaxy , BioMart , EpiGRAPH and UCSC Cancer Genomics Browser. BioMart at this point mostly offers flexible export of user-defined tracks and regions.
Galaxy provides a richer, text-centric suite of operations. EpiGraph presents a solid set of statistical routines focused on analysis of user-defined case-control regions. The recently introduced UCSC Cancer Genomics Browser visualizes clinical omics data, as well as providing patient-centric statistical analyses. We have developed novel statistical methodology and a robust software system for comparative analysis of sequence-level genomic data, enabling integrative systems biology, at the intersection of genomics, computational science and statistics. We focus on inferential investigations, where two genomic annotations, or tracks, are compared in order to find significant deviation from null-model behavior.
Tracks may be defined by the researcher or extracted from the sizable library provided with the system. The system is open-ended, facilitating extensions by the user community. Figure 1 Flow diagram of the mathematics of genomic tracks. Genomic tracks are represented as geometric objects on the line defined by the base pairs of the genome sequence: (unmarked (UP) or marked (MP)) points, (unmarked (US) or marked (MS)) segments, and functions (F). The biologist identifies the two tracks to be compared, and the Genomic HyperBrowser detects their type.
The biological question of interest is stated in terms of mathematical relations between the types of the two tracks. The relevant questions are proposed by the system. The biologist then selects the question and needs to specify the null hypothesis. For this purpose she is called to decide about what structures are preserved in each track, and how to randomize the rest. Thereafter, the Genomic HyperBrowser identifies the relevant test statistics, and computes actual P-values, either exactly or by Monte Carlo testing. Results are then reported, both for a global analysis, answering the question on the whole genome (or area of study), and for a local analysis. Here, the area is divided into bins, and the answer is given per bin.
P-values, test-statistic, and effect sizes are reported, as tables and graphics. Significance is reported when found, after correction for multiple testing. Abstract representation of genomic elements A genome annotation track is a collection of objects of a specific genomic feature, such as genes, with base-pair-specific locations from the start of chromosome 1 to the end of chromosome Y. Tracks vary in biological content, but also in the form of the information they contain. A track representing genes contains positional information that can be reduced to 'segments' (intervals of base pairs) along the genome. A track of SNPs can be reduced to points (single base pairs) on the genome.
The expression values of a gene, or the alleles of a SNP, are non-positional information parts and are attributed as 'marks' (numerical or categorical) to the corresponding positional objects, that is, segments or points. Finally, a track of DNA melting assigns a temperature to each base pair, describing a 'function' on the genome. We thus define five genomic types: unmarked points (UP), marked points (MP), unmarked segments (US), marked segments (MS) and functions (F). These five types completely represent every one-dimensional geometry with marks. Catalogue of investigations We translate biological hypotheses of interest into a study of mathematical relations between genomic tracks, leading to a large collection of possible generic investigations.
Consider the relation between histone modifications and gene expression, as investigated by visual inspection in (Figure S1 in Additional file ). The question is whether the number of nucleosomes with a given histone modification (represented as type UP), counted in a region around the transcription start site (TSS) of a gene, correlates with the expression of the gene.
The second track is represented as marked segments (MS). This study of histone modifications and gene expressions can then be phrased as a generic investigation between a pair of tracks (T1, T2) of type UP and MS: are the number of T1 points inside T2 segments correlated with T2 marks?
Figure shows the results when repeating this analysis for all histone modifications studied in , and different regions around the TSS. Micro focus net express 4.0 free download. See Section 1 in Additional file for a more detailed example investigation, analyzing the genome coverage by different gene definitions. Figure 2 Gene regulation by histone modifications. The correlation between occupancy of 21 different histone modifications and gene expression within 4 different regions around the TSS (up- and downstream, 1 and 20 kb), sorted by correlation in 1-kb upstream regions. Sixteen of 21 histone modifications show significant correlation in 1-kb upstream regions, while inspection of the actual value of Kendall's tau (Table S1 in Additional file ) shows very little effect size for 6 of these 16 (.
To illustrate the value of local analysis, we consider viral integration events in the human genome. These may result in disease and may also be a consequence of retroviral gene therapy. examined integration for six types of retroviruses, with different viral integrases, thus having different integration sites (type UP). Using these data, we asked whether there are hotspots of integration inside 2-kb flanking regions of predicted promoters (type US), that is, whether and where the points are falling inside the segments more than expected by chance. Figure displays the hotspots as calculated P-values in bins across the genome, using the subset of murine leukemia virus (MLV) sites.
We find locations of increased integration, thus generating hypotheses on the role of integration site sequences and their context. Figure 3 Viral integration sites.
Plot of false discovery rate (FDR)-adjusted P-values along the genome, in 30-Mbp bins. Small P-values indicate regions where murine leukemia virus (MLV) integrates inside 2-kb regions around FirstEF promoters more frequently than by chance. The FDR cutoff at 10% is shown as a dashed line. Bhakta prahlada story in telugu pdf. The inset of a local area (chromosome 1:153,250,001-153,450,000) indicates FirstEF promoters expanded by 2 kb in both directions, MLV integration sites, RefSeq genes, and unflanked FirstEF sites.
Local analysis may be used to avoid drawing incorrect conclusions from global investigations. Consider the repressive histone modification H3K27me3 as studied in.
Data from ChIP-chip experiments on mouse chromosome 17 were analyzed, finding that H3K27me3 falls in domains that are enriched in short interspersed nuclear element (SINE) and depleted in long interspersed nuclear element (LINE) repeats. Using the line of enquiry raised in , we asked whether H3K27me3 regions (type US) significantly overlap with SINE repeats (type US), but here using formal statistical testing at the base pair level. The chosen null model only allows local rearrangements of genomic elements (for more detail, see next section). This preserves local biological structure, but allows for some controlled level of randomness. Performing this test globally on the whole chromosome 17 leads to rejection of the null hypothesis ( P = 10 -4), in line with. However, a local analysis leads to a deeper understanding. At a 5-Mbp scale, no significant findings were obtained in any of the 19 bins (10% false discovery rate (FDR)-corrected).
The frequency of H3K27me3 segments varies considerably along chromosome 17 (Figure S2 in Additional file ), which may cause the observed discrepancy between local and global results. Precise specification of null models A crucial aspect of an investigation is the precise formalization of the null model, which should reflect the combination of stochastic and selective events that constitutes the evolution behind the observed genomic feature. Consider again the example of H3K27me3 versus repeating elements. In the chosen null model, we preserved the repeat segments exactly, but permuted the positions of the H3K27me3 segments, while preserving segment and intersegment lengths. We then computed the total overlap between the segments, and used a Monte Carlo test to quantify the departure from the null model. The effect of using alternative null models is shown in Table.
The null model examined in the first column, which does not preserve the dependency between neighboring base pairs, produces lower P-values. Unrealistically simple null models may thus lead to false positives. In fact, two simulated independent tracks may appear to have a significant association if their individual characteristics are not appropriately modeled (Section 2 in Additional file ). In this example, the choice between the biologically more reasonable null models is difficult.
The two other columns of Table include models that preserve more of the biological structure. The fact that these models do not lead to clear rejection of the null hypotheses suggests that we in this case lack strong evidence against the null hypothesis.
Thus, examining the results obtained for a set of different null models may often contribute important information. The null model should reflect biological realism, but also allow sufficient variation to permit the construction of tests. A set of simulated synthetic tracks is provided as an aid for assessing appropriate null models (Additional file ). The number of significant bins of the overlap test between H3K27me3 segments and SINE repeats under different preservation and randomization rules for the null model. The test was performed in 19 bins on mouse chromosome 17, with the MEFB1 cell line. (Use of the MEFF cell line gave similar results; Table S2 in Additional file ).
In this case, less preservation of biological structure leads to smaller P-values. Also, randomizing the SINE track gave smaller P-values than randomizing the H3K27me3 track (or both). The Genomic HyperBrowser allows the user to define an appropriate null model by specifying (a) a preservation rule for each track, and (b) a stochastic process, describing how the non-preserved elements should be randomized. Preservation fixes elements or characteristics of a track as present in the data. For each genomic type, we have developed a hierarchy of less and less strict preservation rules, starting from preserving the entire track exactly (Section 3 in Additional file ). For example, these preservation options for unmarked segments can be assumed: (i) preserve all, as in data; (ii) preserve segments and intervals between segments, in number and length, but not their ordering; (iii) preserve only the segments, in number and length, but not their position; (iv) preserve only the number of base pairs in segments, not segment position or number.
Depending on the test statistic T, the level of preservation and the chosen randomization, P-values are computed exactly, asymptotically or by standard or sequential Monte Carlo ,. Confounder tracks The relation between two tracks of interest may often be modulated by a third track. Such a third track may act as a confounder, leading, if ignored, to dubious conclusions on the relation between the two tracks of interest. Consider the relation of coding regions to the melting stability of the DNA double helix.
Melting forks have been found to coincide with exon boundaries –. Although few studies have reported statistical measures of such correlation , the correlation is confirmed by a straightforward investigation. Tracks (type F) representing the probabilities of melting fork locations in Saccharomyces cerevisiae, were compared to tracks containing all exon boundaries (Figure ). We asked if the melting fork probabilities (P) were higher than expected at the exon boundaries (E) than elsewhere. In the null model, the function was conserved, while points were uniformly randomized in each chromosome. Monte Carlo testing was carried out on the chromosomes separately, giving P-values.
Figure 4 Comparison of exon boundary locations and melting fork probability peaks. Independent analyses were carried out on left and right exon boundaries as compared to left- and right-facing melting forks, respectively. In the upper part, dashed vertical lines indicate left (L, red) and right (R, blue) exon boundaries.
In the lower part, probabilities of left- and right-facing melting forks appear as red and blue peaks, respectively. The black curve shows the GC content in a 100-bp sliding window (values on right axis). An alternative view is that the GC content, being higher inside exons than outside, contains information about exon location that is simply carried over, or decoded, by a melting analysis, thus acting as a confounder. We have developed a methodology to investigate such situations further.
Non-preserved elements of a null model can be randomized according to a non-homogeneous Poisson process with a base-pair-varying intensity, which can depend on a third (or several) modulating genomic tracks ,. We have defined an algebra for the construction of intensities, where tracks are combined, to allow rich and flexible constructions of randomness (see Materials and methods). To investigate the influence of GC content on the exon-melting relation, we first generated a pair of custom tracks (type F), assigning to each base the value given by the GC content in the 100-bp left and right flanking regions, respectively, weighted by a linearly decreasing function. These two functions were used, together with the exon boundary track, to create an intensity curve proportional to the probability of exon points, given GC content (see Materials and methods). When performing the same analysis as before, but now using the null model based on this intensity curve (rather than assuming uniformity), a significant relationship was found in only one yeast chromosome (Table S3 in Additional file ). In conclusion, there is a melting-exon relationship in yeast, but it may simply be a consequence of differences in GC content at the exon boundaries (high GC inside, low GC outside), which may exist for biological reasons not involving melting fork locations. Resolving complexity: system architecture The Genomic HyperBrowser is an integrated, open-source system for genome analysis.
It is continually evolving, supporting 28 different analyses for significance testing, as well as 62 different descriptive statistics. The system currently hosts 184,500 tracks. Most of these represent literature-based information, previously mostly utilized in network-based approaches.
As natural language based text mining allows for the identification of a wide variety of biological entities, we have generated tracks representing genomic locations associated with terms for the complete gene ontology tree, all Medical Subject Heading (MeSH) terms, chemicals, and anatomy. The system is implemented in Python , a high-level programming language that allows fast and robust software development. A main weakness of Python compared to languages like C is its slower performance. Thus, a two-level architecture has been designed. At the highest level, Python objects and logic have been used extensively to provide the required flexibility. At the base-pair level, data are handled as low-level vectors, combining near-optimal storage with efficient indexing, allowing the use of vector operations to ensure speed. Interoperability with standard file formats in the field is provided by parallel storage of original file formats and preprocessed vector representations.
To reduce the memory footprint of analyses on genome-wide data, an iterative divide-and-conquer algorithm is automatically carried out when applicable. A further speedup is achieved by memoizing intermediate results to disk, automatically retrieving them when needed for the same or different analyses on the same track(s) at any subsequent time, by any user.
The system provides a web-based user interface with a low entry point. However, the complex interdependencies between the large body of available tracks, a number of syntactically different analyses, and a range of choices for constructing null models, all pose challenges to the concepts of simplicity and ease of use. In order to simplify the task of making choices, a step-wise approach has been implemented, displaying only the relevant options at each stage. This guided approach hides unnecessary complexities from the researcher, while confronting her with important design choices as needed. We rely on a dynamic system to infer appropriate options, aiding maintenance. The list of selectable tracks is based on scans of available files on disk.
The list of relevant questions is based on short runs of all implemented analyses, using a minimal part of the actual data from the selected tracks. For each analysis, a set of relevant options is defined. The dynamics of the system also provides automatic removal of analyses that fail to run, enhancing system robustness. Allowing extensibility along with efficiency and system dynamics is a challenge. The complexities of the software solutions are hidden in the backbone of the system, simplifying coding of statistical modules.
Free Online Anatomy Atlas
Each module declares the data types it supports and which results are needed from other modules. The backbone automatically checks whether the selected tracks meet the requirements, and if so, makes sure the intermediate computations are carried out in correct order. Redundant computations are avoided through the use of a RAM-based memoization scheme. The system also provides a component-based framework for Monte Carlo tests, where any test statistic can be combined with any relevant randomization algorithm, simplifying development. In addition, a framework for writing unit and integration tests is included. Further details on the system architecture are provided in Section 4 in Additional file.
Step-by-step guide to HyperBrowser analysis One of the main goals of the Genomic HyperBrowser is to facilitate sophisticated statistical analyses. A range of textual guides and screencasts are available in the help section at the web page, demonstrating execution of various analyses, how to work with private data, and more. To give an impression of the user experience, we here provide a step-by-step guide to the analysis of broad local enrichment (BLOC) segments versus SINE repeats, as discussed in the section on 'Precise specification of null models'.
First, we open 'hyperbrowser.uio.no' in a web browser and we select the 'Perform analysis' tool under 'The Genomic HyperBrowser' in the left-hand menu. We select the mouse genome (mm8) and continue to select tracks of interest. As the first track, we select 'Chromatin'-'Histone modifications'-'BLOC segments'-'MEFB1'. These are the BLOC segments according to the algorithm of Pauler et al.
for the MEFB1 cell line. As the second track, we select 'Sequence'-'Repeating elements'-'SINE'. Now that both tracks have been selected, a list of relevant investigations is presented in the interface (that is, investigations that are compatible with the genomic types of the two tracks: US versus US). We select the question of 'Overlap?'
In the 'Hypothesis testing' category, and the options relevant for this analysis are subsequently displayed in the interface. The different choices for 'Null model' will produce the various numbers in Table (six different choices are directly available from the list.
The other variants can be achieved by reversing the selection order of the tracks). The original BLOC paper focused on chromosome 17. We want to perform a local analysis along this chromosome, avoiding the first three megabases that are centromeric. Under 'Region and scale' we thus choose to 'Compare in' a custom specified region, writing 'chr17:3m-' as 'Region of the genome' and writing '5 m' (5 megabases) as 'Bin size'. Clicking the 'Start analysis' button will then perform an appropriate statistical test according to the selected null model assumption, and output textual and graphical results to a new Galaxy history element.
Figure shows the user interface covering all selections above and Figure shows the answer page that results from this analysis. Figure 5 Screenshots of the Genomic HyperBrowser.
(a) Screenshot of the main interface for selecting analysis options. The selections for the example relating H3K27me3 BLOCs to SINE repeats have been pre-selected. In the interface, the user selects a genome build followed by two tracks. A list of relevant investigations is then presented, based on the genomic types of the two tracks. After selecting an investigation, the interface presents the user with a choice of null models, alternative hypotheses and other relevant options. (b) Screenshot of the results of the analysis. The question asked by the user is presented at the top, in this case: 'Are 'MEFB1 (BLOC segments)' overlapping 'SINE (Repeating elements)' more than expected by chance?'
A first, simplistic answer is then presented: 'No support from data for this conclusion in any bin'. A more precise answer follows, detailing any global P-values, a summary of local FDR-corrected P-values, the particular set of null and alternative hypotheses tested, in addition to a legend of the test statistic that has been used. Further links to a PDF file containing the statistical details of the test, and to more detailed tables of relevant statistics for both the global and the local analysis are also included. The global result table also includes links to plots and export opportunities for the individual statistics.
This example assumed the BLOC segments were already in the system. If not, they could simply be uploaded to the Galaxy history and then selected in the first track menu as '- From history (bed, wig) -'-'your BLOC history element'. For information on how to use the Galaxy system, we refer to the Galaxy web site. The current leap in high-throughput sequencing technology is opening the way for a range of genome-wide annotations beyond the presently abundant gene-centric data.
Not least, chromatin-related data are becoming increasingly important for understanding higher-level organization and regulation of the genome. As is typical for a subfield that has not reached maturation, analysis of new massive sequence-level data is performed on a per-project basis. For instance, a paper on the ENCODE project describes how inference can be done by Monte Carlo testing, sampling bins for one of the real tracks at random genome locations under the null hypothesis. Independently, a newer study of histone modifications instead permuted bins of data for one of the tracks. Although genomic visualization tools have been available for several years, few generic tools exist for inference at the sequence level. The following aspects distinguish our work from currently available systems. First, we focus on genomic information of a sequential nature, that is, with specific base-pair locations on a genome, and thus not restricted to only genes.
Second, it focuses on the comparison of pairs of genomic tracks, possibly taking others into account through the concept of intensity tracks. Third, all comparisons are performed using formal statistical testing. Fourth, we provide analyses on any scale, from genome-wide studies to miniature investigations on particular loci. Fifth, we offer flexible choices of null models for exploration and choice where relevant. Finally, we provide a user interface where the user describes the data and the null models, while the system based on this chooses the appropriate statistical test. Comparing this to the EpiGRAPH and Galaxy frameworks, which we believe are the closest existing systems, we find that both require substantial technical expertise when choosing the correct analysis and options.
EpiGRAPH is focused on a specific type of scenario that, according to our cataloguing, amounts to the comparison of unmarked points or segments versus categorically marked segments (with mark being case or control). Galaxy provides a simple user interface, is rich in tools for manipulating and analyzing datasets of diverse formats, but has little support for formal statistical testing. Note also that our system is tightly connected to Galaxy and can make use of all the tools provided within Galaxy. We provide tools for abstraction and cataloguing of what we believe are typical questions of broad interest. The abstractions of genomic data, the proposing of prototype investigations, and the careful attention given to null models simplifies statistical inference for a range of possible research topics.
Our approach invites researchers to build relevant null models in a controlled manner, so that specific biological assumptions can be realistically represented by preservation, randomness and intensity based confounders. In addition, time used for repetitive tasks like file parsing and calculation of descriptive statistics may be significantly reduced. Our system is highly extensible. The software is open source, inviting the community to add new investigations and tools. Attention has been given to component-based coding and simple interfaces, facilitating extensions of the system. The highly specialized nature of many research investigations poses a major challenge for a generic system such as the one presented here.
Even though a range of analyses and options are provided, chances are that at a given level of complexity, functionality beyond what is provided by a generic system will be needed. Still, the time and effort used to reach such a point may be shortened considerably, and it should in many cases be possible to meet demands through custom extensions.
Genomic mechanisms commonly involve more than two tracks, and the current focus on pair-wise interrogations is limiting. Our methodology allows the incorporation of additional tracks through the concept of an intensity track that modulates the null hypothesis, acting as a confounder. However, the investigation of genuine multi-track interactions is not yet possible within the system, as complex modeling and testing of multiple dependencies will be required. Attention should be given to the trade-off between fine resolution and lack of precision. When large bins are considered, there may be too little homogeneity, while small bins may contain too little data. There is also an unresolved trade-off relating to preservation of tracks in null-hypotheses construction: too little preservation may give unrealistically small P-values, while too strong preservation may give too limited randomness. On a more specific note, a set of tissue-specific analytical options would be beneficial with respect to many types of experimental data - for example, chromatin, expression and also gene subset tracks.
Such options are now under development. Novel sequencing technologies are instrumental in realizing the personalized genomes , and with them the task of identifying phenotype-associated information contained in each genome.
An imminent challenge in understanding cellular organization is that of the three dimensions of the genome. While a number of genomes have been sequenced, and a number of important cellular elements have been mapped on a linear scale, the mapping of the three-dimensional organization of the DNA and chromatin in the nucleus is still only in its beginnings. Consequently, the impact of this organization on cell regulation is still largely unresolved. However, the advent of methods like Hi-C permits detailed maps of three-dimensional DNA interactions to be combined with coarser methods of mapping of other elements. It appears that looking simultaneously at multiple scales seems important for understanding the dynamics of different functional aspects, from chromosomal domains down to the nucleosome scale. The need for taking multiple scales into account has recently been emphasized in both theoretical and analytical settings ,. Consequently, statistical genomics needs to consider several scales when proper analytical routines are developed.
Our approach is open to three-dimensional extensions, where the bins, which are flexibly selected in the system, will become three-dimensional volumes, and local comparison will be within each volume. What appears much more complex is the level of dependence of such volumes.
But as the three-dimensional organization of the genome will become increasingly known, appropriate volume topologies will be possible, so that neighboring volumes representing three-dimensional contiguity may be used as a basis for statistical tests. By introducing a generic methodology to genome analysis, we find that a range of genomic data sets can be represented by the same mathematical objects, and that a small set of such objects suffice to describe the bulk of current data sets. Similarly, a range of biological investigations can be reduced to similar statistical analyses. The need for precise control of assumptions and other parameters can furthermore be met by generic concepts such as preservation and randomization, local analysis (binning) and confounder tracks.
Applying these ideas on a sample set of genomic investigations underlines that the generic concepts fit naturally to concrete analyses, and that such a generic treatment may expose vagueness of biological conclusions or expose unforeseen issues. A re-analysis of the relation between BLOC segments of histone modification and SINE repeats shows that conclusions regarding direct overlap at the base-pair level depends on the randomizations used in the significance analysis. Using biologically reasonable null models, the correspondence between BLOC segments and SINE repeats appears not to be due to overlap at the base-pair level, but rather seems to be due to local variation in intensities of both tracks. This does not directly oppose the original conclusions, but brings further insight into the nature of the relation. Similarly, an analysis of the relation between DNA melting and exon location confirms the conclusion from previous studies that exon boundaries coincide with gradients of melting temperature. However, taking GC content into account as a possible confounder, the analysis does not suggest a direct functional relation between melting and exons. Instead, it suggests that the association is due to the relationship of both exons and melting tracks to GC content.
We believe the generic concepts and challenges identified by our work will trigger community efforts to improve genome analysis methodology. The Genomic HyperBrowser demonstrates the feasibility of applying our approach to large-scale genomic datasets, providing a concrete basis for further research and development in inferential genomics. We thus consider the solutions presented here more like a start than an end of this important endeavor. Statistical methods A track is defined over the whole genome or only in parts of it, masking away the rest. In a local analysis, statistical tests are performed in each bin with sufficient sample size.
Resizing of bins allows for localization of events (similarities, differences, and so on, between the two tracks) with flexible precision. Preservation rules leads to conditional P-values that are not necessarily ordered, even if the preservation mechanism is incremental.
Statistical tests have been tried on simulated data, also when model assumptions are not completely fulfilled. Standard Monte Carlo requires deciding on the number of Monte Carlo samples. We suggest at least two to five times the number of tests, in order to allow for FDR adjustment. Additionally, we adopt sequential Monte Carlo, where the algorithm continues sampling until the observed statistic has been exceeded a given number of times (say 20).
This gives better estimates of small P-values with overall reduced computations. Intensity tracks are used to define non-standard null hypothesis. Several strategies for building intensity curves are described in Section 3 in Additional file. Intensity curves allow performing randomizations that mimic another track (or a combination of tracks), useful to account for confounding effects. For unmarked points, the intensity curve can be any regular function λ 0(b) where b is the position along, say, a chromosome. If λ 0(b) = c (constant), points are uniformly distributed. As another example, λ 0(b) can be a kernel density estimate based on the track of observed points.
In general, the intensity λ 0(b) may depend on several different tracks g 1, g 2., g k, through a function s, so that λ 0(b) = s(g 1(b), g 2(b)., g k(b)), for example, λ 0(b) = c + Σβ ig i(b). An important case that requires a special choice of intensity track is when the comparison between two tracks T 1 and T 2 might be confounded by a third, confounder, track T 3. This is discussed in further detail in Section 5 in Additional file for the melting-exon example, where each track depends on a function of the GC content. Software system The Genomic HyperBrowser is implemented in Python , version 2.7. It runs as a stand-alone application tightly connected to the Galaxy framework , using the version dated 2010-10-04.
The user interface is based on Mako templates for Python , version 0.2.5, and Javascript library Jquery , version 1.4.2. The software uses NumPy , version 1.5.1rc1, for disk based vector mapping and fast vector operations. R , version 2.10.1, is used for plotting and basic statistical routines, using the RPy API , version 1.0.3. The software is open source and freely available, using GPL version 3, and can be downloaded from. The Genomic HyperBrowser runs on a dedicated Linux server, with large computations offloaded to the Titan cluster.
Comments are closed.
|
AuthorWrite something about yourself. No need to be fancy, just an overview. ArchivesCategories |