TRAPT

A novel deep learning framework for transcription regulators prediction via integraing large-scale epigenomic data.

Analysis

Jump to analysis results.

Please enter your Task ID.
Mean Reciprocal Rank
TRAPT algorithm pipeline
A ' A ' A ... 0.5 0.1 0.6 0.3 Gene Epigenomic Signals PredictionAUC Intergration Epigenomic Signals Intergration 1.000.5 Weight TSS±100 kb around TSS of gene DHS region 20kb TR1TR2TR3TR4Other Activity score ... TR Epigenomic Signals Prediction TRs TRs TRs Eip-samples Genes KNN EncoderDecoder FeatureMaps GraphEncoder GraphDecoder μ TRs TRs TRs LabeledDocumentSet GUnlabeledDocumentSet G’ 1110000 ...... H3K27ac Genes ... Rank bycorrelation Sparsity Group Lasso Features Genes ATAC ... 0.2 0.5 0.4 0.6 Group by 10 group1groupn GenesH3K27acATAC Genes Teacher modelStudent model RP Matrix Teacher modelStudent modelNN modelTR-RP MatrixTR-Epigenomic SignalsGene-Epigenomic Signals GenesGenes TRsTRs GenesGenes TR-Epigenomic Activity Matrix ... 0.6 0.5 0.3 0.1 ... 0.6 0.5 0.4 0.2 ... group1group2groupn 1.000.5 Weight TSS±100kb around TSS of gene DHS region 20kb TR1TR2 ... TR-RP modelEpigenomic-RP model AS ∑ i M j = 1 AUC ij AUC j = SampleAUC TR1TR2TR3TR4Other0.80.90.30.6 ... SampleAUC TR1TR2TR3TR4Other0.70.90.20.6 ... RP Matrix select samples correlation 10 samples genes Query genes TR-Epigenomic Signals TRs Genes Aggregate A A ' A ' group2 TRs Genes TRs Genes Gene-Epigenomic Signals TRs TR TR ...... distilled knowledge distilled knowledge σ μ g σ g Z g h h g Omics Labels xczc Eip-samples A T A C d b A T A C d b S E S E d b d b G e n e C R E T F G e n e C R E T c o F G e n e C R E C R TcoF Base TcoF Base TF TG TF TG CR db CR db FeatureMaps Eip-samples Eip-samples Eip-samples Eip-sample 1Eip-sample 2Eip-sample 3Eip-sample n Epigenomics