search

calico / basenji

Sequential regulatory activity predictions with deep convolutional neural networks.
Apache License 2.0
400 stars 121 forks source link

norm the effect size across chromosomes #121

Open yinhongwei4079 opened 2 years ago

yinhongwei4079 commented 2 years ago

@davek44 we had predicted the effect size of single snp in single chromosome by basenji_sad . we checked the scrpit by basenji and found the the basenji_sad_norm and we want to if we had to norm the effect sizes of snps for all chr by using the script?

davek44 commented 2 years ago

I never quite finished integrating basenji_sad_norm into a complete workflow. If you are planning to compare scores across tasks (corresponding to separate experiments), then it can be useful to normalize the scores. The raw scores will be influenced by uninteresting factors like sequencing depth.

This is easiest when you are making predictions for a large number of SNPs, but it's harder when you are predicting a single SNP since you'd need to precompute the normalization parameters. Is that the scenario that you're in?