yjqiu
commented
3 years ago @rmd13 We make predictions by comparing deltaSVM scores with the experimental thresholds as defined by median value of pbSNPs identified in SNP-SELEX. The absolute scale of the deltaSVM score is NOT informative especially compared across TFs.
Yes. Gain means 2nd allele is stronger than 1st. Take SNP chr1_100_A_G as an example, gain means TF binds G stronger than A.
The bind score is predicted by gkm-SVM which measures the binding strength of the entire sequence. But difference between gkm-SVM is not used to measure allelic binding and it is only used to measure if the sequence can be bind by the TF. Allelic binding is measured by deltaSVM scores.
I run the deltaSVM online using demo 3 SNPs provided online server as example. In the results table there are 2 columes: "deltaSVM_score" and "preferred_allele".Sometimes there are SNP-TF lines shows relative high delta_SVM value but the "preferred_allele" shows "None", while there are relative a little bit lower "deltaSVM_score" but "preferred_allele" shows "Gain". How to interpret this scenario? Was the "preferred_allele" determined not only by delta_SVM score but also by the score for each alternative allele？ Does the "Gain" means allelel2 bind stronger to SNP than allele1? But sometims allele2_bind is -0.706563 but allele1_bind is 0.562300, lookes the reverse but still shows 'Gain'. Thanks.