cg
view grant.html @ 43:8cce366da1e5
.
author | bshanks@bshanks.dyndns.org |
---|---|
date | Wed Apr 15 00:50:34 2009 -0700 (16 years ago) |
parents | 282ba15dcfbe |
children | c4a887af9b0b |
line source
1 Specific aims
2 Massivenew datasets obtained with techniques such as in situ hybridization (ISH), immunohistochemistry, or in situ trans-
3 genic reporter allow the expression levels of many genes at many locations to be compared. Our goal is to develop automated
4 methods to relate spatial variation in gene expression to anatomy. We want to find marker genes for specific anatomical
5 regions, and also to draw new anatomical maps based on gene expression patterns. We have three specific aims:
6 (1) develop an algorithm to screen spatial gene expression data for combinations of marker genes which selectively target
7 anatomical regions
8 (2) develop an algorithm to suggest new ways of carving up a structure into anatomical regions, based on spatial patterns
9 in gene expression
10 (3) create a 2-D “flat map” dataset of the mouse cerebral cortex that contains a flattened version of the Allen Mouse
11 Brain Atlas ISH data, as well as the boundaries of cortical anatomical areas. This will involve extending the functionality of
12 Caret, an existing open-source scientific imaging program. Use this dataset to validate the methods developed in (1) and (2).
13 In addition to validating the usefulness of the algorithms, the application of these methods to cerebral cortex will produce
14 immediate benefits, because there are currently no known genetic markers for many cortical areas. The results of the project
15 will support the development of new ways to selectively target cortical areas, and it will support the development of a
16 method for identifying the cortical areal boundaries present in small tissue samples.
17 All algorithms that we develop will be implemented in an open-source software toolkit. The toolkit, as well as the
18 machine-readable datasets developed in aim (3), will be published and freely available for others to use.
19 Background and significance
20 Aim 1
21 Machine learning terminology: supervised learning
22 The task of looking for marker genes for anatomical regions means that one is looking for a set of genes such that, if the
23 expression level of those genes is known, then the locations of the regions can be inferred.
24 If we define the regions so that they cover the entire anatomical structure to be divided, then instead of saying that we
25 are using gene expression to find the locations of the regions, we may say that we are using gene expression to determine to
26 which region each voxel within the structure belongs. We call this a classification task, because each voxel is being assigned
27 to a class (namely, its region).
28 Therefore, an understanding of the relationship between the combination of their expression levels and the locations of
29 the regions may be expressed as a function. The input to this function is a voxel, along with the gene expression levels
30 within that voxel; the output is the regional identity of the target voxel, that is, the region to which the target voxel belongs.
31 We call this function a classifier. In general, the input to a classifier is called an instance, and the output is called a label
32 (or a class label).
33 The object of aim 1 is not to produce a single classifier, but rather to develop an automated method for determining a
34 classifier for any known anatomical structure. Therefore, we seek a procedure by which a gene expression dataset may be
35 analyzed in concert with an anatomical atlas in order to produce a classifier. Such a procedure is a type of a machine learning
36 procedure. The construction of the classifier is called training (also learning), and the initial gene expression dataset used
37 in the construction of the classifier is called training data.
38 In the machine learning literature, this sort of procedure may be thought of as a supervised learning task, defined as a
39 task in which the goal is to learn a mapping from instances to labels, and the training data consists of a set of instances
40 (voxels) for which the labels (regions) are known.
41 Each gene expression level is called a feature, and the selection of which genes1 to include is called feature selection.
42 Feature selection is one component of the task of learning a classifier. Some methods for learning classifiers start out with
43 a separate feature selection phase, whereas other methods combine feature selection with other aspects of training.
44 One class of feature selection methods assigns some sort of score to each candidate gene. The top-ranked genes are then
45 chosen. Some scoring measures can assign a score to a set of selected genes, not just to a single gene; in this case, a dynamic
46 procedure may be used in which features are added and subtracted from the selected set depending on how much they raise
47 the score. Such procedures are called “stepwise” or “greedy”.
48 Although the classifier itself may only look at the gene expression data within each voxel before classifying that voxel, the
49 learning algorithm which constructs the classifier may look over the entire dataset. We can categorize score-based feature
50 selection methods depending on how the score of calculated. Often the score calculation consists of assigning a sub-score to
51 each voxel, and then aggregating these sub-scores into a final score (the aggregation is often a sum or a sum of squares). If
52 only information from nearby voxels is used to calculate a voxel’s sub-score, then we say it is a local scoring method. If only
53 information from the voxel itself is used to calculate a voxel’s sub-score, then we say it is a pointwise scoring method.
54 Key questions when choosing a learning method are: What are the instances? What are the features? How are the
55 features chosen? Here are four principles that outline our answers to these questions.
56 Principle 1: Combinatorial gene expression It is too much to hope that every anatomical region of interest will be
57 identified by a single gene. For example, in the cortex, there are some areas which are not clearly delineated by any gene
58 included in the Allen Brain Atlas (ABA) dataset. However, at least some of these areas can be delineated by looking at
59 combinations of genes (an example of an area for which multiple genes are necessary and sufficient is provided in Preliminary
60 Results). Therefore, each instance should contain multiple features (genes).
61 Principle 2: Only look at combinations of small numbers of genes When the classifier classifies a voxel, it is
62 only allowed to look at the expression of the genes which have been selected as features. The more data that is available to
63 a classifier, the better that it can do. For example, perhaps there are weak correlations over many genes that add up to a
64 strong signal. So, why not include every gene as a feature? The reason is that we wish to employ the classifier in situations
65 in which it is not feasible to gather data about every gene. For example, if we want to use the expression of marker genes as
66 a trigger for some regionally-targeted intervention, then our intervention must contain a molecular mechanism to check the
67 expression level of each marker gene before it triggers. It is currently infeasible to design a molecular trigger that checks the
68 level of more than a handful of genes. Similarly, if the goal is to develop a procedure to do ISH on tissue samples in order
69 to label their anatomy, then it is infeasible to label more than a few genes. Therefore, we must select only a few genes as
70 features.
71 Principle 3: Use geometry in feature selection
72 _________________________________________
73 1Strictly speaking, the features are gene expression levels, but we’ll call them genes.
74 When doing feature selection with score-based methods, the simplest thing to do would be to score the performance of
75 each voxel by itself and then combine these scores (pointwise scoring). A more powerful approach is to also use information
76 about the geometric relations between each voxel and its neighbors; this requires non-pointwise, local scoring methods. See
77 Preliminary Results for evidence of the complementary nature of pointwise and local scoring methods.
78 Principle 4: Work in 2-D whenever possible
79 There are many anatomical structures which are commonly characterized in terms of a two-dimensional manifold. When
80 it is known that the structure that one is looking for is two-dimensional, the results may be improved by allowing the analysis
81 algorithm to take advantage of this prior knowledge. In addition, it is easier for humans to visualize and work with 2-D
82 data.
83 Therefore, when possible, the instances should represent pixels, not voxels.
84 Related work
85 There is a substantial body of work on the analysis of gene expression data, however, most of this concerns gene expression
86 data which is not fundamentally spatial.
87 As noted above, there has been much work on both supervised learning and there are many available algorithms for
88 each. However, the algorithms require the scientist to provide a framework for representing the problem domain, and the
89 way that this framework is set up has a large impact on performance. Creating a good framework can require creatively
90 reconceptualizing the problem domain, and is not merely a mechanical “fine-tuning” of numerical parameters. For example,
91 we believe that domain-specific scoring measures (such as gradient similarity, which is discussed in Preliminary Work) may
92 be necessary in order to achieve the best results in this application.
93 We are aware of three existing efforts to find marker genes using spatial gene expression data using automated methods.
94 [? ] describes GeneAtlas. GeneAtlas allows the user to construct a search query by freely demarcating one or two 2-D
95 regions on sagittal slices, and then to specify either the strength of expression or the name of another gene whose expression
96 pattern is to be matched. GeneAtlas differs from our Aim 1 in at least two ways. First, GeneAtlas finds only single genes,
97 whereas we will also look for combinations of genes2. Second, at least for the custom spatial search, Gene Atlas appears to
98 use a simple pointwise scoring method (strength of expression), whereas we will also use geometric metrics such as gradient
99 similarity.
100 [2 ] describes AGEA, ”Anatomic Gene Expression Atlas”. AGEA has three components:
101 * Gene Finder: The user selects a seed voxel and the system (1) chooses a cluster which includes the seed voxel, (2)
102 yields a list of genes which are overexpressed in that cluster. (note: the ABA website also contains pre-prepared lists of
103 overexpressed genes for selected structures)
104 * Correlation: The user selects a seed voxel and the shows the user how much correlation there is between the gene
105 expression profile of the seed voxel and every other voxel.
106 * Clusters: AGEA includes a precomputed hierarchial clustering of voxels based on a recursive bifurcation algorithm
107 with correlation as the similarity metric.
108 Gene Finder is different from our Aim 1 in at least three ways. First, Gene Finder finds only single genes, whereas we
109 will also look for combinations of genes. Second, gene finder can only use overexpression as a marker, whereas we will also
110 search for underexpression. Third, Gene Finder uses a simple pointwise score3, whereas we will also use geometric scores
111 such as gradient similarity. The Preliminary Data section contains evidence that each of our three choices is the right one.
112 In summary, none of the previous projects explores combinations of marker genes, and none of their publications compare
113 the results obtained by using different algorithms or scoring methods.
114 Aim 2
115 Machine learning terminology: clustering
116 If one is given a dataset consisting merely of instances, with no class labels, then analysis of the dataset is referred to as
117 unsupervised learning in the jargon of machine learning. One thing that you can do with such a dataset is to group instances
118 together. A set of similar instances is called a cluster, and the activity of finding grouping the data into clusters is called
119 clustering or cluster analysis.
120 The task of deciding how to carve up a structure into anatomical regions can be put into these terms. The instances are
121 once again voxels (or pixels) along with their associated gene expression profiles. We make the assumption that voxels from
122 the same region have similar gene expression profiles, at least compared to the other regions. This means that clustering
123 voxels is the same as finding potential regions; we seek a partitioning of the voxels into regions, that is, into clusters of voxels
124 with similar gene expression.
125 _________________
126 2See Preliminary Data for an example of an area which cannot be marked by any single gene in the dataset, but which can be marked by a
127 combination.
128 3“Expression energy ratio”, which captures overexpression.
129 It is desirable to determine not just one set of regions, but also how these regions relate to each other, if at all; perhaps
130 some ofthe regions are more similar to each other than to the rest, suggesting that, although at a fine spatial scale they
131 could be considered separate, on a coarser spatial scale they could be grouped together into one large region. This suggests
132 the outcome of clustering may be a hierarchial tree of clusters, rather than a single set of clusters which partition the voxels.
133 This is called hierarchial clustering.
134 Similarity scores
135 A crucial choice when designing a clustering method is how to measure similarity, across either pairs of instances, or
136 clusters, or both. There is much overlap between scoring methods for feature selection (discussed above under Aim 1) and
137 scoring methods for similarity.
138 Spatially contiguous clusters; image segmentation
139 We have shown that aim 2 is a type of clustering task. In fact, it is a special type of clustering task because we have
140 an additional constraint on clusters; voxels grouped together into a cluster must be spatially contiguous. In Preliminary
141 Results, we show that one can get reasonable results without enforcing this constraint, however, we plan to compare these
142 results against other methods which guarantee contiguous clusters.
143 Perhaps the biggest source of continguous clustering algorithms is the field of computer vision, which has produced a
144 variety of image segmentation algorithms. Image segmentation is the task of partitioning the pixels in a digital image into
145 clusters, usually contiguous clusters. Aim 2 is similar to an image segmentation task. There are two main differences; in
146 our task, there are thousands of color channels (one for each gene), rather than just three. There are imaging tasks which
147 use more than three colors, however, for example multispectral imaging and hyperspectral imaging, which are often used
148 to process satellite imagery. A more crucial difference is that there are various cues which are appropriate for detecting
149 sharp object boundaries in a visual scene but which are not appropriate for segmenting abstract spatial data such as gene
150 expression. Although many image segmentation algorithms can be expected to work well for segmenting other sorts of
151 spatially arranged data, some of these algorithms are specialized for visual images.
152 Dimensionality reduction
153 Unlike aim 1, there is no externally-imposed need to select only a handful of informative genes for inclusion in the
154 instances. However, some clustering algorithms perform better on small numbers of features. There are techniques which
155 “summarize” a larger number of features using a smaller number of features; these techniques go by the name of feature
156 extraction or dimensionality reduction. The small set of features that such a technique yields is called the reduced feature
157 set. After the reduced feature set is created, the instances may be replaced by reduced instances, which have as their features
158 the reduced feature set rather than the original feature set of all gene expression levels. Note that the features in the reduced
159 feature set do not necessarily correspond to genes; each feature in the reduced set may be any function of the set of gene
160 expression levels.
161 Another use for dimensionality reduction is to visualize the relationships between regions. For example, one might want
162 to make a 2-D plot upon which each region is represented by a single point, and with the property that regions with similar
163 gene expression profiles should be nearby on the plot (that is, the property that distance between pairs of points in the plot
164 should be proportional to some measure of dissimilarity in gene expression). It is likely that no arrangement of the points on
165 a 2-D plan will exactly satisfy this property – however, dimensionality reduction techniques allow one to find arrangements
166 of points that approximately satisfy that property. Note that in this application, dimensionality reduction is being applied
167 after clustering; whereas in the previous paragraph, we were talking about using dimensionality reduction before clustering.
168 Clustering genes rather than voxels
169 Although the ultimate goal is to cluster the instances (voxels or pixels), one strategy to achieve this goal is to first cluster
170 the features (genes). There are two ways that clusters of genes could be used.
171 Gene clusters could be used as part of dimensionality reduction: rather than have one feature for each gene, we could
172 have one reduced feature for each gene cluster.
173 Gene clusters could also be used to directly yield a clustering on instances. This is because many genes have an expression
174 pattern which seems to pick out a single, spatially continguous region. Therefore, it seems likely that an anatomically
175 interesting region will have multiple genes which each individually pick it out4. This suggests the following procedure:
176 cluster together genes which pick out similar regions, and then to use the more popular common regions as the final clusters.
177 In the Preliminary Data we show that a number of anatomically recognized cortical regions, as well as some “superregions”
178 formed by lumping together a few regions, are associated with gene clusters in this fashion.
179 Related work
180 We are aware of three existing efforts to cluster spatial gene expression data.
181 _________________________________________
182 4This would seem to contradict our finding in aim 1 that some cortical areas are combinatorially coded by multiple genes. However, it is
183 possible that the currently accepted cortical maps divide the cortex into regions which are unnatural from the point of view of gene expression;
184 perhaps there is some other way to map the cortex for which each region can be identified by single genes.
185 [5 ] describes an analysis of the anatomy of the hippocampus using the ABA dataset. In addition to manual analysis,
186 two clustering methods were employed, a modified Non-negative Matrix Factorization (NNMF), and a hierarchial recursive
187 bifurcation clustering scheme based on correlation as the similarity score. The paper yielded impressive results, proving the
188 usefulness of such research. We have run NNMF on the cortical dataset5 and while the results are promising (see Preliminary
189 Data), we think that it will be possible to find an even better method. In addition, this paper described a visual screening
190 of the data, specifically, a visual analysis of 6000 genes with the primary purpose of observing how the spatial pattern of
191 their expression coincided with the regions that had been identified by NNMF. We propose to do this sort of screening
192 automatically, which would yield an objective, quantifiable result, rather than qualitative observations.
193 AGEA’s[2] hierarchial clustering differs from our Aim 2 in at least two ways. First, AGEA uses perhaps the simplest
194 possible similarity score (correlation), and does no dimensionality reduction before calculating similarity. While it is possible
195 that a more complex system will not do any better than this, we believe further exploration of alternative methods of scoring
196 and dimensionality reduction is warranted. Second, AGEA did not look at clusters of genes; in Preliminary Data we have
197 shown that clusters of genes may identify interesting spatial regions such as cortical areas.
198 [? ] todo
199 In summary, although these projects obtained hierarchial clusterings, there has not been much comparison between
200 different algorithms or scoring methods, so it is likely that the best clustering method for this application has not yet been
201 found.
202 Aim 3
203 Background
204 The cortex is divided into areas and layers. To a first approximation, the parcellation of the cortex into areas can
205 be drawn as a 2-D map on the surface of the cortex. In the third dimension, the boundaries between the areas continue
206 downwards into the cortical depth, perpendicular to the surface. The layer boundaries run parallel to the surface. One can
207 picture an area of the cortex as a slice of many-layered cake.
208 Although it is known that different cortical areas have distinct roles in both normal functioning and in disease processes,
209 there are no known marker genes for many cortical areas. When it is necessary to divide a tissue sample into cortical areas,
210 this is a manual process that requires a skilled human to combine multiple visual cues and interpret them in the context of
211 their approximate location upon the cortical surface.
212 Even the questions of how many areas should be recognized in cortex, and what their arrangement is, are still not
213 completely settled. A proposed division of the cortex into areas is called a cortical map. In the rodent, the lack of a
214 single agreed-upon map can be seen by contrasting the recent maps given by Swanson[4] on the one hand, and Paxinos
215 and Franklin[3] on the other. While the maps are certainly very similar in their general arrangement, significant differences
216 remain in the details.
217 The Allen Mouse Brain Atlas dataset
218 The Allen Mouse Brain Atlas (ABA) data was produced by doing in-situ hybridization on slices of male, 56-day-old
219 C57BL/6J mouse brains. Pictures were taken of the processed slice, and these pictures were semi-automatically analyzed
220 in order to create a digital measurement of gene expression levels at each location in each slice. Per slice, cellular spatial
221 resolution is achieved. Using this method, a single physical slice can only be used to measure one single gene; many different
222 mouse brains were needed in order to measure the expression of many genes.
223 Next, an automated nonlinear alignment procedure located the 2D data from the various slices in a single 3D coordinate
224 system. In the final 3D coordinate system, voxels are cubes with 200 microns on a side. There are 67x41x58 = 159,326
225 voxels in the 3D coordinate system, of which 51,533 are in the brain[2].
226 Mus musculus, the common house mouse, is thought to contain about 22,000 protein-coding genes[6]. The ABA contains
227 data on about 20,000 genes in sagittal sections, out of which over 4,000 genes are also measured in coronal sections. Our
228 dataset is derived from only the coronal subset of the ABA, because the sagittal data does not cover the entire cortex,
229 and has greater registration error[2]. Genes were selected by the Allen Institute for coronal sectioning based on, “classes of
230 known neuroscientific interest... or through post hoc identification of a marked non-ubiquitous expression pattern”[2].
231 The ABA is not the only large public spatial gene expression dataset. Other such resources include GENSAT[?],
232 GenePaint[?], its sister project GeneAtlas[?], BGEM[?], EMAGE[?], EurExpress (http://www.eurexpress.org/ee/; Eur-
233 Express data is also entered into EMAGE), todo. With the exception of the ABA, GenePaint, and EMAGE, most of these
234 resources, have not (yet) extracted the expression intensity from the ISH images and registered the results into a single 3-D
235 space, and only ABA and EMAGE make this form of data available for public download from the website. Many of these
236 resources focus on developmental gene expression.
237 Significance
238 ___________________________
239 5We ran “vanilla” NNMF, whereas the paper under discussion used a modified method. Their main modification consisted of adding a soft
240 spatial contiguity constraint. However, on our dataset, NNMF naturally produced spatially contiguous clusters, so no additional constraint was
241 needed. The paper under discussion also mentions that they tried a hierarchial variant of NNMF, which we have not yet tried.
242 The method developed in aim (1) will be applied to each cortical area to find a set of marker genes such that the
243 combinatorial expression pattern of those genes uniquely picks out the target area. Finding marker genes will be useful for
244 drug discovery as well as for experimentation because marker genes can be used to design interventions which selectively
245 target individual cortical areas.
246 The application of the marker gene finding algorithm to the cortex will also support the development of new neuroanatom-
247 ical methods. In addition to finding markers for each individual cortical areas, we will find a small panel of genes that can
248 find many of the areal boundaries at once. This panel of marker genes will allow the development of an ISH protocol that
249 will allow experimenters to more easily identify which anatomical areas are present in small samples of cortex.
250 The method developed in aim (3) will provide a genoarchitectonic viewpoint that will contribute to the creation of
251 a better map. The development of present-day cortical maps was driven by the application of histological stains. It is
252 conceivable that if a different set of stains had been available which identified a different set of features, then the today’s
253 cortical maps would have come out differently. Since the number of classes of stains is small compared to the number of
254 genes, it is likely that there are many repeated, salient spatial patterns in the gene expression which have not yet been
255 captured by any stain. Therefore, current ideas about cortical anatomy need to incorporate what we can learn from looking
256 at the patterns of gene expression.
257 While we do not here propose to analyze human gene expression data, it is conceivable that the methods we propose to
258 develop could be used to suggest modifications to the human cortical map as well.
259 Related work
260 [2 ] describes the application of AGEA to the cortex. The paper describes interesting results on the structure of correlations
261 between voxel gene expression profiles within a handful of cortical areas. However, this sort of analysis is not related to
262 either of our aims, as it neither finds marker genes, nor does it suggest a cortical map based on gene expression data. Neither
263 of the other components of AGEA can be applied to cortical areas; AGEA’s Gene Finder cannot be used to find marker
264 genes for cortical areas; and AGEA’s hierarchial clustering does not produce clusters corresponding to cortical areas.
265 In both cases, the root cause is that pairwise correlations between the gene expression of voxels in different areas but
266 the same layer are stronger than pairwise correlations between the gene expression of voxels in different layers but the same
267 area. Therefore a pairwise voxel correlation clustering algorithm will always create clusters representing cortical layers, not
268 areas. This is why the hierarchial clustering does not find cortical areas6. The reason that Gene Finder cannot find marker
269 genes for cortical areas is that in Gene Finder, although the user chooses a seed voxel, Gene Finder chooses the ROI for
270 which genes will be found, and it creates that ROI by (pairwise voxel correlation) clustering around the seed.
271 In summary, for all three aims, (a) none of the previous projects explores combinations of marker genes, (b) there has
272 been almost no comparison of different algorithms or scoring methods, and (c) there has been no work on computationally
273 finding marker genes for cortical areas, or on finding a hierarchial clustering that will yield a map of cortical areas de novo
274 from gene expression data.
275 Our project is guided by a concrete application with a well-specified criterion of success (how well we can find marker
276 genes for / reproduce the layout of cortical areas), which will provide a solid basis for comparing different methods.
277 _________________________________________
278 6There are clusters which presumably correspond to the intersection of a layer and an area, but since one area will have many layer-area
279 intersection clusters, further work is needed to make sense of these.
280 Preliminary work
281 Format conversion between SEV, MATLAB, NIFTI
282 We have created software to (politely) download all of the SEV files from the Allen Institute website. We have also created
283 software to convert between the SEV, MATLAB, and NIFTI file formats, as well as some of Caret’s file formats.
284 Flatmap of cortex
285 We downloaded the ABA data and applied a mask to select only those voxels which belong to cerebral cortex. We divided
286 the cortex into hemispheres.
287 Using Caret[1], we created a mesh representation of the surface of the selected voxels. For each gene, for each node of
288 the mesh, we calculated an average of the gene expression of the voxels “underneath” that mesh node. We then flattened
289 the cortex, creating a two-dimensional mesh.
290 We sampled the nodes of the irregular, flat mesh in order to create a regular grid of pixel values. We converted this grid
291 into a MATLAB matrix.
292 We manually traced the boundaries of each cortical area from the ABA coronal reference atlas slides. We then converted
293 these manual traces into Caret-format regional boundary data on the mesh surface. We projected the regions onto the 2-d
294 mesh, and then onto the grid, and then we converted the region data into MATLAB format.
295 At this point, the data is in the form of a number of 2-D matrices, all in registration, with the matrix entries representing
296 a grid of points (pixels) over the cortical surface:
297 ∙A 2-D matrix whose entries represent the regional label associated with each surface pixel
298 ∙For each gene, a 2-D matrix whose entries represent the average expression level underneath each surface pixel
299 We created a normalized version of the gene expression data by subtracting each gene’s mean expression level (over all
300 surface pixels) and dividing each gene by its standard deviation.
301 The features and the target area are both functions on the surface pixels. They can be referred to as scalar fields over
302 the space of surface pixels; alternately, they can be thought of as images which can be displayed on the flatmapped surface.
303 To move beyond a single average expression level for each surface pixel, we plan to create a separate matrix for each
304 cortical layer to represent the average expression level within that layer. Cortical layers are found at different depths in
305 different parts of the cortex. In preparation for extracting the layer-specific datasets, we have extended Caret with routines
306 that allow the depth of the ROI for volume-to-surface projection to vary.
307 In the Research Plan, we describe how we will automatically locate the layer depths. For validation, we have manually
308 demarcated the depth of the outer boundary of cortical layer 5 throughout the cortex.
309 Feature selection and scoring methods
310 Correlation Recall that the instances are surface pixels, and consider the problem of attempting to classify each instance
311 as either a member of a particular anatomical area, or not. The target area can be represented as a binary mask over the
312 surface pixels.
313 One class of feature selection scoring method are those which calculate some sort of “match” between each gene image
314 and the target image. Those genes which match the best are good candidates for features.
315 One of the simplest methods in this class is to use correlation as the match score. We calculated the correlation between
316 each gene and each cortical area.
317 todo: fig
318 Conditional entropy An information-theoretic scoring method is to find features such that, if the features (gene
319 expression levels) are known, uncertainty about the target (the regional identity) is reduced. Entropy measures uncertainty,
320 so what we want is to find features such that the conditional distribution of the target has minimal entropy. The distribution
321 to which we are referring is the probability distribution over the population of surface pixels.
322 The simplest way to use information theory is on discrete data, so we discretized our gene expression data by creating,
323 for each gene, five thresholded binary masks of the gene data. For each gene, we created a binary mask of its expression
324 levels using each of these thresholds: the mean of that gene, the mean minus one standard deviation, the mean minus two
325 standard deviations, the mean plus one standard deviation, the mean plus two standard deviations.
326 Now, for each region, we created and ran a forward stepwise procedure which attempted to find pairs of gene expression
327 binary masks such that the conditional entropy of the target area’s binary mask, conditioned upon the pair of gene expression
328 binary masks, is minimized.
329 This finds pairs of genes which are most informative (at least at these discretization thresholds) relative to the question,
330 “Is this surface pixel a member of the target area?”.
334 Figure 1: The top row shows the three genes which (individually) best predict area AUD, according to logistic regression.
335 The bottom row shows the three genes which (individually) best match area AUD, according to gradient similarity. From
336 left to right and top to bottom, the genes are Ssr1, Efcbp1, Aph1a, Ptk7, Aph1a again, and Lepr
337 todo: fig
338 Gradient similarity We noticed that the previous two scoring methods, which are pointwise, often found genes whose
339 pattern of expression did not look similar in shape to the target region. Fort his reason we designed a non-pointwise local
340 scoring method to detect when a gene had a pattern of expression which looked like it had a boundary whose shape is similar
341 to the shape of the target region. We call this scoring method “gradient similarity”.
342 One might say that gradient similarity attempts to measure how much the border of the area of gene expression and
343 the border of the target region overlap. However, since gene expression falls off continuously rather than jumping from its
344 maximum value to zero, the spatial pattern of a gene’s expression often does not have a discrete border. Therefore, instead
345 of looking for a discrete border, we look for large gradients. Gradient similarity is a symmetric function over two images
346 (i.e. two scalar fields). It is is high to the extent that matching pixels which have large values and large gradients also have
347 gradients which are oriented in a similar direction. The formula is:
348 ∑
349 pixel<img src="cmsy7-32.png" alt="∈" />pixels cos(abs(∠∇1 -∠∇2)) ⋅|∇1| + |∇2|
350 2 ⋅ pixel_value1 + pixel_value2
351 2
352 where ∇1 and ∇2 are the gradient vectors of the two images at the current pixel; ∠∇i is the angle of the gradient of
353 image i at the current pixel; |∇i| is the magnitude of the gradient of image i at the current pixel; and pixel_valuei is the
354 value of the current pixel in image i.
355 The intuition is that we want to see if the borders of the pattern in the two images are similar; if the borders are similar,
356 then both images will have corresponding pixels with large gradients (because this is a border) which are oriented in a
357 similar direction (because the borders are similar).
358 Gradient similarity provides information complementary to correlation
359 To show that gradient similarity can provide useful information that cannot be detected via pointwise analyses, consider
360 Fig. . The top row of Fig. displays the 3 genes which most match area AUD, according to a pointwise method7. The bottom
361 row displays the 3 genes which most match AUD according to a method which considers local geometry8 The pointwise
362 method in the top row identifies genes which express more strongly in AUD than outside of it; its weakness is that this
363 includes many areas which don’t have a salient border matching the areal border. The geometric method identifies genes
364 whose salient expression border seems to partially line up with the border of AUD; its weakness is that this includes genes
365 which don’t express over the entire area. Genes which have high rankings using both pointwise and border criteria, such as
366 Aph1a in the example, may be particularly good markers. None of these genes are, individually, a perfect marker for AUD;
367 we deliberately chose a “difficult” area in order to better contrast pointwise with geometric methods.
368 Combinations of multiple genes are useful
369 Here we give an example of a cortical area which is not marked by any single gene, but which can be identified combi-
370 natorially. according to logistic regression, gene wwc19 is the best fit single gene for predicting whether or not a pixel on
371 _________________________________________
372 7For each gene, a logistic regression in which the response variable was whether or not a surface pixel was within area AUD, and the predictor
373 variable was the value of the expression of the gene underneath that pixel. The resulting scores were used to rank the genes in terms of how well
374 they predict area AUD.
375 8For each gene the gradient similarity (see section ??) between (a) a map of the expression of each gene on the cortical surface and (b) the
376 shape of area AUD, was calculated, and this was used to rank the genes.
377 9“WW, C2 and coiled-coil domain containing 1”; EntrezGene ID 211652
381 Figure 2: Upper left: wwc1. Upper right: mtif2. Lower left: wwc1 + mtif2 (each pixel’s value on the lower left is the sum
382 of the corresponding pixels in the upper row). Within each picture, the vertical axis roughly corresponds to anterior at the
383 top and posterior at the bottom, and the horizontal axis roughly corresponds to medial at the left and lateral at the right.
384 The red outline is the boundary of region MO. Pixels are colored approximately according to the density of expressing cells
385 underneath each pixel, with red meaning a lot of expression and blue meaning little.
386 the cortical surface belongs to the motor area (area MO). The upper-left picture in Figure shows wwc1’s spatial expression
387 pattern over the cortex. The lower-right boundary of MO is represented reasonably well by this gene, however the gene
388 overshoots the upper-left boundary. This flattened 2-D representation does not show it, but the area corresponding to the
389 overshoot is the medial surface of the cortex. MO is only found on the lateral surface (todo).
390 Gene mtif210 is shown in figure the upper-right of Fig. . Mtif2 captures MO’s upper-left boundary, but not its lower-right
391 boundary. Mtif2 does not express very much on the medial surface. By adding together the values at each pixel in these
392 two figures, we get the lower-left of Figure . This combination captures area MO much better than any single gene.
393 Areas which can be identified by single genes
394 todo
395 Underexpression of a gene can serve as a marker
396 todo
397 Specific to Aim 1 (and Aim 3)
398 Forward stepwise logistic regression todo
399 SVM on all genes at once
400 In order to see how well one can do when looking at all genes at once, we ran a support vector machine to classify cortical
401 surface pixels based on their gene expression profiles. We achieved classification accuracy of about 81%11. As noted above,
402 however, a classifier that looks at all the genes at once isn’t practically useful.
403 The requirement to find combinations of only a small number of genes limits us from straightforwardly applying many
404 of the most simple techniques from the field of supervised machine learning. In the parlance of machine learning, our task
405 combines feature selection with supervised learning.
406 Decision trees
407 todo
408 Specific to Aim 2 (and Aim 3)
409 Raw dimensionality reduction results
410 todo
411 (might want to incld nnMF since mentioned above)
412 _________________________________________
413 10“mitochondrial translational initiation factor 2”; EntrezGene ID 76784
414 115-fold cross-validation.
415 Dimensionality reduction plus K-means or spectral clustering
416 Many areas are captured by clusters of genes
417 todo
418 todo
419 Research plan
420 Further work on flatmapping
421 In anatomy, the manifold of interest is usually either defined by a combination of two relevant anatomical axes (todo),
422 or by the surface of the structure (as is the case with the cortex). In the former case, the manifold of interest is a plane, but
423 in the latter case it is curved. If the manifold is curved, there are various methods for mapping the manifold into a plane.
424 In the case of the cerebral cortex, it remains to be seen which method of mapping the manifold into a plane is optimal
425 for this application. We will compare mappings which attempt to preserve size (such as the one used by Caret[1]) with
426 mappings which preserve angle (conformal maps).
427 Although there is much 2-D organization in anatomy, there are also structures whose shape is fundamentally 3-dimensional.
428 If possible, we would like the method we develop to include a statistical test that warns the user if the assumption of 2-D
429 structure seems to be wrong.
430 todo amongst other things:
431 Develop algorithms that find genetic markers for anatomical regions
432 1.Develop scoring measures for evaluating how good individual genes are at marking areas: we will compare pointwise,
433 geometric, and information-theoretic measures.
434 2.Develop a procedure to find single marker genes for anatomical regions: for each cortical area, by using or combining
435 the scoring measures developed, we will rank the genes by their ability to delineate each area.
436 3.Extend the procedure to handle difficult areas by using combinatorial coding: for areas that cannot be identified by any
437 single gene, identify them with a handful of genes. We will consider both (a) algorithms that incrementally/greedily
438 combine single gene markers into sets, such as forward stepwise regression and decision trees, and also (b) supervised
439 learning techniques which use soft constraints to minimize the number of features, such as sparse support vector
440 machines.
441 4.Extend the procedure to handle difficult areas by combining or redrawing the boundaries: An area may be difficult
442 to identify because the boundaries are misdrawn, or because it does not “really” exist as a single area, at least on the
443 genetic level. We will develop extensions to our procedure which (a) detect when a difficult area could be fit if its
444 boundary were redrawn slightly, and (b) detect when a difficult area could be combined with adjacent areas to create
445 a larger area which can be fit.
446 Apply these algorithms to the cortex
447 1.Create open source format conversion tools: we will create tools to bulk download the ABA dataset and to convert
448 between SEV, NIFTI and MATLAB formats.
449 2.Flatmap the ABA cortex data: map the ABA data onto a plane and draw the cortical area boundaries onto it.
450 3.Find layer boundaries: cluster similar voxels together in order to automatically find the cortical layer boundaries.
451 4.Run the procedures that we developed on the cortex: we will present, for each area, a short list of markers to identify
452 that area; and we will also present lists of “panels” of genes that can be used to delineate many areas at once.
453 Develop algorithms to suggest a division of a structure into anatomical parts
454 1.Explore dimensionality reduction algorithms applied to pixels: including TODO
455 2.Explore dimensionality reduction algorithms applied to genes: including TODO
456 3.Explore clustering algorithms applied to pixels: including TODO
457 4.Explore clustering algorithms applied to genes: including gene shaving, TODO
458 5.Develop an algorithm to use dimensionality reduction and/or hierarchial clustering to create anatomical maps
459 6.Run this algorithm on the cortex: present a hierarchial, genoarchitectonic map of the cortex
460 Bibliography & References Cited
461 [1]D C Van Essen, H A Drury, J Dickson, J Harwell, D Hanlon, and C H Anderson. An integrated software suite for surface-
462 based analyses of cerebral cortex. Journal of the American Medical Informatics Association: JAMIA, 8(5):443–59, 2001.
463 PMID: 11522765.
464 [2]Lydia Ng, Amy Bernard, Chris Lau, Caroline C Overly, Hong-Wei Dong, Chihchau Kuan, Sayan Pathak, Susan M Sunkin,
465 Chinh Dang, Jason W Bohland, Hemant Bokil, Partha P Mitra, Luis Puelles, John Hohmann, David J Anderson, Ed S
466 Lein, Allan R Jones, and Michael Hawrylycz. An anatomic gene expression atlas of the adult mouse brain. Nat Neurosci,
467 12(3):356–362, March 2009.
468 [3]George Paxinos and Keith B.J. Franklin. The Mouse Brain in Stereotaxic Coordinates. Academic Press, 2 edition, July
469 2001.
470 [4]Larry Swanson. Brain Maps: Structure of the Rat Brain. Academic Press, 3 edition, November 2003.
471 [5]Carol L. Thompson, Sayan D. Pathak, Andreas Jeromin, Lydia L. Ng, Cameron R. MacPherson, Marty T. Mortrud,
472 Allison Cusick, Zackery L. Riley, Susan M. Sunkin, Amy Bernard, Ralph B. Puchalski, Fred H. Gage, Allan R. Jones,
473 Vladimir B. Bajic, Michael J. Hawrylycz, and Ed S. Lein. Genomic anatomy of the hippocampus. Neuron, 60(6):1010–
474 1021, December 2008.
475 [6]Robert H Waterston, Kerstin Lindblad-Toh, Ewan Birney, Jane Rogers, Josep F Abril, Pankaj Agarwal, Richa Agarwala,
476 Rachel Ainscough, Marina Alexandersson, Peter An, Stylianos E Antonarakis, John Attwood, Robert Baertsch, Jonathon
477 Bailey, Karen Barlow, Stephan Beck, Eric Berry, Bruce Birren, Toby Bloom, Peer Bork, Marc Botcherby, Nicolas Bray,
478 Michael R Brent, Daniel G Brown, Stephen D Brown, Carol Bult, John Burton, Jonathan Butler, Robert D Campbell,
479 Piero Carninci, Simon Cawley, Francesca Chiaromonte, Asif T Chinwalla, Deanna M Church, Michele Clamp, Christopher
480 Clee, Francis S Collins, Lisa L Cook, Richard R Copley, Alan Coulson, Olivier Couronne, James Cuff, Val Curwen, Tim
481 Cutts, Mark Daly, Robert David, Joy Davies, Kimberly D Delehaunty, Justin Deri, Emmanouil T Dermitzakis, Colin
482 Dewey, Nicholas J Dickens, Mark Diekhans, Sheila Dodge, Inna Dubchak, Diane M Dunn, Sean R Eddy, Laura Elnitski,
483 Richard D Emes, Pallavi Eswara, Eduardo Eyras, Adam Felsenfeld, Ginger A Fewell, Paul Flicek, Karen Foley, Wayne N
484 Frankel, Lucinda A Fulton, Robert S Fulton, Terrence S Furey, Diane Gage, Richard A Gibbs, Gustavo Glusman, Sante
485 Gnerre, Nick Goldman, Leo Goodstadt, Darren Grafham, Tina A Graves, Eric D Green, Simon Gregory, Roderic Guig,
486 Mark Guyer, Ross C Hardison, David Haussler, Yoshihide Hayashizaki, LaDeana W Hillier, Angela Hinrichs, Wratko
487 Hlavina, Timothy Holzer, Fan Hsu, Axin Hua, Tim Hubbard, Adrienne Hunt, Ian Jackson, David B Jaffe, L Steven
488 Johnson, Matthew Jones, Thomas A Jones, Ann Joy, Michael Kamal, Elinor K Karlsson, Donna Karolchik, Arkadiusz
489 Kasprzyk, Jun Kawai, Evan Keibler, Cristyn Kells, W James Kent, Andrew Kirby, Diana L Kolbe, Ian Korf, Raju S
490 Kucherlapati, Edward J Kulbokas, David Kulp, Tom Landers, J P Leger, Steven Leonard, Ivica Letunic, Rosie Levine, Jia
491 Li, Ming Li, Christine Lloyd, Susan Lucas, Bin Ma, Donna R Maglott, Elaine R Mardis, Lucy Matthews, Evan Mauceli,
492 John H Mayer, Megan McCarthy, W Richard McCombie, Stuart McLaren, Kirsten McLay, John D McPherson, Jim
493 Meldrim, Beverley Meredith, Jill P Mesirov, Webb Miller, Tracie L Miner, Emmanuel Mongin, Kate T Montgomery,
494 Michael Morgan, Richard Mott, James C Mullikin, Donna M Muzny, William E Nash, Joanne O Nelson, Michael N
495 Nhan, Robert Nicol, Zemin Ning, Chad Nusbaum, Michael J O’Connor, Yasushi Okazaki, Karen Oliver, Emma Overton-
496 Larty, Lior Pachter, Gens Parra, Kymberlie H Pepin, Jane Peterson, Pavel Pevzner, Robert Plumb, Craig S Pohl, Alex
497 Poliakov, Tracy C Ponce, Chris P Ponting, Simon Potter, Michael Quail, Alexandre Reymond, Bruce A Roe, Krishna M
498 Roskin, Edward M Rubin, Alistair G Rust, Ralph Santos, Victor Sapojnikov, Brian Schultz, Jrg Schultz, Matthias S
499 Schwartz, Scott Schwartz, Carol Scott, Steven Seaman, Steve Searle, Ted Sharpe, Andrew Sheridan, Ratna Shownkeen,
500 Sarah Sims, Jonathan B Singer, Guy Slater, Arian Smit, Douglas R Smith, Brian Spencer, Arne Stabenau, Nicole Stange-
501 Thomann, Charles Sugnet, Mikita Suyama, Glenn Tesler, Johanna Thompson, David Torrents, Evanne Trevaskis, John
502 Tromp, Catherine Ucla, Abel Ureta-Vidal, Jade P Vinson, Andrew C Von Niederhausern, Claire M Wade, Melanie Wall,
503 Ryan J Weber, Robert B Weiss, Michael C Wendl, Anthony P West, Kris Wetterstrand, Raymond Wheeler, Simon
504 Whelan, Jamey Wierzbowski, David Willey, Sophie Williams, Richard K Wilson, Eitan Winter, Kim C Worley, Dudley
505 Wyman, Shan Yang, Shiaw-Pyng Yang, Evgeny M Zdobnov, Michael C Zody, and Eric S Lander. Initial sequencing and
506 comparative analysis of the mouse genome. Nature, 420(6915):520–62, December 2002. PMID: 12466850.
508 _______________________________________________________________________________________________________
509 stuff i dunno where to put yet (there is more scattered through grant-oldtext):
510 Principle 4: Work in 2-D whenever possible
511 —
512 note:
513 do we need to cite: no known markers, impressive results?
514 two hemis
515 “genomic anatomy” is a name found in the titles of one of the cited papers which seems good