1.1 --- a/grant.html Mon Apr 13 03:20:00 2009 -0700 1.2 +++ b/grant.html Mon Apr 13 03:21:04 2009 -0700 1.3 @@ -300,16 +300,12 @@ 1.4 think that more automation of the parts that this paper’s authors did manually 1.5 will be possible). 1.6 and [?] describes AGEA. todo 1.7 - In the Preliminary Work, we show that 1.8 - The creation of a domain-specific scoring measure may be required in order 1.9 - to achieve good performance, and it is not impossible that the algorithms them- 1.10 - selves will have to be extended. We plan to test out existing algorithms and 1.11 - scoring measures, 1.12 - Therefore, we anticipate 1.13 - Therefore, it is unclear which of the 1.14 - todo 1.15 - vs. AGEA – i wrote something on this but i’m going to rewrite it 1.16 -__________________________ 1.17 + Preliminary work 1.18 + Format conversion between SEV, MATLAB, NIFTI 1.19 + todo 1.20 + Flatmap of cortex 1.21 + todo 1.22 +_______________________ 1.23 2We ran “vanilla” NNMF, whereas the paper under discussion used a modified method. 1.24 Their main modification consisted of adding a soft spatial contiguity constraint. However, 1.25 on our dataset, NNMF naturally produced spatially contiguous clusters, so no additional 1.26 @@ -318,11 +314,6 @@ 1.27 not any more impressive than the results of the non-hierarchial variant. 1.28 7 1.29 1.30 - Preliminary work 1.31 - Format conversion between SEV, MATLAB, NIFTI 1.32 - todo 1.33 - Flatmap of cortex 1.34 - todo 1.35 Using combinations of multiple genes is necessary and sufficient to 1.36 delineate some cortical areas 1.37 Here we give an example of a cortical area which is not marked by any 1.38 @@ -352,7 +343,15 @@ 1.39 genes which express more strongly in AUD than outside of it; its weakness is that 1.40 this includes many areas which don’t have a salient border matching the areal 1.41 border. The geometric method identifies genes whose salient expression border 1.42 -__________________________ 1.43 + seems to partially line up with the border of AUD; its weakness is that this 1.44 + includes genes which don’t express over the entire area. Genes which have high 1.45 + rankings using both pointwise and border criteria, such as Aph1a in the example, 1.46 + may be particularly good markers. None of these genes are, individually, a 1.47 + perfect marker for AUD; we deliberately chose a “difficult” area in order to 1.48 + better contrast pointwise with geometric methods. 1.49 + Areas which can be identified by single genes 1.50 + todo 1.51 +____________________ 1.52 3“WW, C2 and coiled-coil domain containing 1”; EntrezGene ID 211652 1.53 4“mitochondrial translational initiation factor 2”; EntrezGene ID 76784 1.54 5For each gene, a logistic regression in which the response variable was whether or not a 1.55 @@ -374,8 +373,6 @@ 1.56 the boundary of region MO. Pixels are colored approximately according to the 1.57 density of expressing cells underneath each pixel, with red meaning a lot of 1.58 expression and blue meaning little. 1.59 - 9 1.60 - 1.61 1.62 1.63 Figure 2: The top row shows the three genes which (individually) best predict 1.64 @@ -383,14 +380,8 @@ 1.65 genes which (individually) best match area AUD, according to gradient similar- 1.66 ity. From left to right and top to bottom, the genes are Ssr1, Efcbp1, Aph1a, 1.67 Ptk7, Aph1a again, and Lepr 1.68 - seems to partially line up with the border of AUD; its weakness is that this 1.69 - includes genes which don’t express over the entire area. Genes which have high 1.70 - rankings using both pointwise and border criteria, such as Aph1a in the example, 1.71 - may be particularly good markers. None of these genes are, individually, a 1.72 - perfect marker for AUD; we deliberately chose a “difficult” area in order to 1.73 - better contrast pointwise with geometric methods. 1.74 - Areas which can be identified by single genes 1.75 - todo 1.76 + 9 1.77 + 1.78 Specific to Aim 1 (and Aim 3) 1.79 Forward stepwise logistic regression todo 1.80 SVM on all genes at once 1.81 @@ -405,10 +396,6 @@ 1.82 our task combines feature selection with supervised learning. 1.83 Decision trees 1.84 todo 1.85 -____________________ 1.86 - 75-fold cross-validation. 1.87 - 10 1.88 - 1.89 Specific to Aim 2 (and Aim 3) 1.90 Raw dimensionality reduction results 1.91 todo 1.92 @@ -432,6 +419,10 @@ 1.93 with a handful of genes. We will consider both (a) algorithms that incre- 1.94 mentally/greedily combine single gene markers into sets, such as forward 1.95 stepwise regression and decision trees, and also (b) supervised learning 1.96 +__________________________ 1.97 + 75-fold cross-validation. 1.98 + 10 1.99 + 1.100 techniques which use soft constraints to minimize the number of features, 1.101 such as sparse support vector machines. 1.102 4. Extend the procedure to handle difficult areas by combining or redrawing 1.103 @@ -447,8 +438,6 @@ 1.104 LAB formats. 1.105 2. Flatmap the ABA cortex data: map the ABA data onto a plane and draw 1.106 the cortical area boundaries onto it. 1.107 - 11 1.108 - 1.109 3. Find layer boundaries: cluster similar voxels together in order to auto- 1.110 matically find the cortical layer boundaries. 1.111 4. Run the procedures that we developed on the cortex: we will present, for 1.112 @@ -468,8 +457,9 @@ 1.113 clustering to create anatomical maps 1.114 6. Run this algorithm on the cortex: present a hierarchial, genoarchitectonic 1.115 map of the cortex 1.116 -______________________________________________ 1.117 - stuff i dunno where to put yet (there is more scattered through grant- 1.118 + 11 1.119 + 1.120 + _______________________________________________________________________________________________________ stuff i dunno where to put yet (there is more scattered through grant- 1.121 oldtext): 1.122 Principle 4: Work in 2-D whenever possible 1.123 In anatomy, the manifold of interest is usually either defined by a combina- 1.124 @@ -489,13 +479,11 @@ 1.125 assumption of 2-D structure seems to be wrong. 1.126 if we need citations for aim 3 significance, http://www.sciencedirect. 1.127 com/science?_ob=ArticleURL&_udi=B6WSS-4V70FHY-9&_user=4429&_coverDate= 1.128 +12%2F26%2F2008&_rdoc=1&_fmt=full&_orig=na&_cdi=7054&_docanchor=&_acct= 1.129 +C000059602&_version=1&_urlVersion=0&_userid=4429&md5=551eccc743a2bfe6e992eee0c3194203# 1.130 +app2 has examples of genetic targeting to specific anatomical regions 1.131 + — 1.132 + note: 1.133 12 1.134 1.135 - 12%2F26%2F2008&_rdoc=1&_fmt=full&_orig=na&_cdi=7054&_docanchor=&_acct= 1.136 - C000059602&_version=1&_urlVersion=0&_userid=4429&md5=551eccc743a2bfe6e992eee0c3194203# 1.137 - app2 has examples of genetic targeting to specific anatomical regions 1.138 - — 1.139 - note: 1.140 - 13 1.141 - 1.142 - 1.143 +

2.1 Binary file grant.odt has changed

3.1 Binary file grant.pdf has changed

4.1 --- a/grant.txt Mon Apr 13 03:20:00 2009 -0700 4.2 +++ b/grant.txt Mon Apr 13 03:21:04 2009 -0700 4.3 @@ -150,18 +150,6 @@ 4.4 and \cite{ng_anatomic_2009} describes AGEA. todo 4.5 4.6 4.7 -In the Preliminary Work, we show that 4.8 - 4.9 -The creation of a domain-specific scoring measure may be required in order to achieve good performance, and it is not impossible that the algorithms themselves will have to be extended. We plan to test out existing algorithms and scoring measures, 4.10 - 4.11 -Therefore, we anticipate 4.12 - 4.13 -Therefore, it is unclear which of the 4.14 - 4.15 -todo 4.16 - 4.17 -vs. AGEA -- i wrote something on this but i'm going to rewrite it 4.18 - 4.19 4.20 == Preliminary work == 4.21