Rui:Genelist
Purpose[edit]
- To organize neuronal gene lists either from literature or from data analysis for everyone
- To provide a short description and guidance for exacting useful information
- To select candidate genes for experimental validation such as FISH or in situ sequencing
ABI ISH 1000[edit]
Highlights[edit]
- Online database profiling ∼1,000 genes in human cortex at cellular resolution
- Diverse expression patterns delineate regional boundaries and distinct cell types
- While 5% of genes vary among human individuals, 21% differ between human and mouse
- Species-specific molecular signatures for major cortical cell types are identified
Description and links[edit]
- Reference link [1]
- Dr. Zhang outlines how he chose the genes for the previous probe set here which includes the ABI Brain Atlas data [2]
Gene list[edit]
Rizi's gene list on 1000 BA8 neuronal single nuclei[edit]
Highlights[edit]
- Nine groups were identified
- Top 279 genes identified that can separate the 9 clusters
Analysis description and links[edit]
- Sample filtering, highly variant gene detection, hierarchical clustering, Random Forest feature selection described here
File:Rizi Clusters 10-22-2014.png
Gene list[edit]
- Criteria to choose top 279 genes
- Nine clusters were identified based on hierarchical clustering analysis of over-dispersed genes
- Clusters were confirmed by random forests algorithm with OOB of 4.69%
Rui's gene list on 1000 BA8 neuronal single nuclei[edit]
Highlights[edit]
- 12 groups in a hierarchical order by a bi-branching cluster strategy
- 400 most differentially expressed genes were identified to separate 2 clusters in each round.
Analysis description and links[edit]
We collected 1064 single neuronal nucleus isolated from human brain Brodmann area 8. After a quality filter based on the difference between zero- and single-nucleus samples, 986 single nuclei were qualified to continue with clustering analysis. To identify neuornal cell types among these 986 samples, we measured similarity among expression profiles of all individual nuclei. The heat map reflecting correlation of each individual to all the others were used to guide for clustering. Correlations revealed two major nuclei population, which was differed significantly in genes expressed in GABAergic inhibitory neurons (GAD1/GAD2 etc.). The separation of these two clusters were further confirmed by K-means clustering and principle component analysis (PCA). We therefore performed a series of unbiased clustering approach that allows hierarchical bi-branching of samples based on gene expression profiles. By doing this, local rather than global gene expression pattern can be associated and subsets of nuclei can be identified based on genes with similar expression. All together, we divided 986 single-nucleus samples into 12 groups and identified 193 signature genes highly expressed in each defined group.
File:12groups.jpgFile:400genes comp.jpg
Gene list[edit]
- Criteria to choose full genes
Starting from 986 single nuclei, we cluster these samples into two groups based on similarity of their gene expression. The most differentially expressed genes (p<0.05) causing such sample separation were identified, among which genes with a fold change at least 4 were chosen as "marker genes". Samples were kept clustering in such bi-branching strategy until 12 groups were identified. Altogether, 315 marker genes were identified responsible for each dividing.
Genes are listed with log2(fold change) and ClusterID. For example, CBLN2's cluster ID is A1/2/3/4/5, which means it supports the first separation between A (A1/2/3/4/5) and B (B1/2/3/4/5/6/7) groups and up-regulated in A group.
Among these 315 genes, they are 68 genes showing up in multiple separations (2-4). Some of the duplicate genes make perfect sense, such as BTBD11 showing up in B1/2/3/4/5/6/7, B1/2/3 and B1 cluster. However, some of them are not easily explain.
Gene lists: media:genelist_Rui_upload_9.19.14.xlsx "HC_Gene" - list all genes based on separations, up-regulated level is presented as red; down-regulated level is in green. "duplicate_unique" - I separate 68 duplicated genes from 247 unique genes which only show up once.
- Criteria to choose group specific genes
Among 247 unique genes, I grab those identified at the very final separations. So there are 152 genes showing up-regulation in one of the 12 groups A1, A2, A3, A4, A5, B1, B2, B3, B4, B5, B6 and B7.
Gene lists: media:genelist_Rui_upload_9.19.14.xlsx "12groupsGene" -
Final gene list (~500): Combined list of Rizi & Rui for in situ sequencing (Matt & Hosuk)[edit]
Highlights[edit]
- Gene Lists from Rizi's and Rui's analyses were combined into one single share-able Google doc
- Gene fields were updated to include functional marker information
Analysis description and links[edit]
- Rizi's Analysis: Obtained top 279 genes ranked in order of importance for ability to identify her 9 distinct clusters.
- Rui's Analysis: Obtained the top 315 genes underlying each cluster separation [3].
- Updated gene information for both lists using ABI's ISH Data[4]
Combined Gene list[edit]
- Criteria to choose genes for in situ Sequencing
- Rizi's Analysis: 279 Unique Genes
- Rui's analyses: 212 Unique Genes not included in Rizi's list
- Genes beginning with MT-* were excluded (mitochondrial in origin)
- Final number = 490 Candidate Genes + 20 Glial markers from the 1000 ABI ISH list
- Gene lists:
TopDiffExpGenes_1000hNucDataset_10282014 media:TopDiffExpGenes_1000hNucDataset_10282014.xlsx
Updated Gene List for Padlock Probes Tab
- Editable Google Doc for this and upcoming gene lists:
TopDiffExpGenes_1000hNucDataset [5]
Overlapped 58 genes suggested for FISH (Dan)[edit]
- Criteria for picking on these 58 genes
Among Rizi's top 100 genes, 58 genes are overlapped with my gene list. Therefore, these 58 genes can be the top candidates for Dan in RNA FISH.
media:genelist_overlaped_9.19.14.xlsx