Hang Xu Stanford University seqFish amp scRNA integration Question What is the minimal number of genes needed for data integration seqFish amp scRNA integration Question What is the minimal number of genes needed for data integration ID: 1044292
Download Presentation The PPT/PDF document "CORTEX seq-FISH: gene selection" is the property of its rightful owner. Permission is granted to download and print the materials on this web site for personal, non-commercial use only, and to display it on your personal computer provided you do not modify the materials and that you retain all copyright notices contained in the materials. By downloading content from our website, you accept the terms of this agreement.
1. CORTEX seq-FISH: gene selectionHang XuStanford University
2. seqFish & scRNA integrationQuestion:What is the minimal number of genes needed for data integration?
3. seqFish & scRNA integrationQuestion:What is the minimal number of genes needed for data integration?What is the minimal number of genes needed to distinguish cell types in both datasets ?
4. Qian et al. 2018, Figure 1 c,dseqFish & scRNA integrationQuestion:What is the minimal number of genes needed for data integration?43 genes !What is the minimal number of genes needed to distinguish cell types in both datasets ?8 major cell types
5. Identify highly variable genesscRNA read countsFilter cells: Num genes >= 200 & < 10000mitochondrial gene percent < 15%Filter genes: min_cells =3 NormalizationAnnotate highly variable genesMin_mean = 0.0125Max_mean = 3Min_dispersion = 0.5Package: Scanpy
6. Identify highly variable genesscRNA read countsFilter cells: Num genes >= 200 & < 10000mitochondrial gene percent < 15%Filter genes: min_cells =3 NormalizationAnnotate highly variable genesMin_mean = 0.0125Max_mean = 3Min_dispersion = 0.5Package: ScanpyAmong the 113 genes overlapped between scRNAseq and seqFish46 genes are annotated as high variable genes
7. Colinear relationship between 46 highly variable genes in scRNAseq data
8. StrategySVM classification modelscRNAseq data
9. StrategySVM classification modelFit modelGain the importance of every featureRemove the least important featureSpecific number of features reachedyesnoRecursiveFeatureElimination(RFE)Cross validation for scoringParameter C: [1e-6, 1e-3, 1]scRNAseq data
10. Feature selectionC = 1e-6f1 weighted
11. Feature selectionC = 1e-6C = 1e-3C = 1e-6C = 1e-3f1 weighted
12. Feature selectionf1 weightedC = 1e-6C = 1e-3C = 1e-6C = 1e-3C = 1C = 1
13. Feature selectionf1 weightedC = 1e-6C = 1e-3C = 1e-6C = 1e-3C = 1C = 112 features7 features
14. scRNAseqFishPrediction resultsAveragez-scoreC: 1e-3Num_features: 12
15. Prediction resultsC: 1e-6Num_features: 7scRNAseqFishAveragez-score
16. Thank you!Questions?
17. C: 1e-3Num_features: 12C: 1e-6Num_features: 7Compare with Original prediction