Extended Data Fig. 3: Nonlinearities and noise across DNNs and TFs.
From: Interpreting cis-regulatory mechanisms from genomic deep neural networks using surrogate models
Examples of the GE nonlinearities and heteroscedastic noise models inferred in the SQUID (GE) analyses performed for Fig. 3. Each plot shows results for a representative sequence among the n = 50 sequences analyzed for each combination of DNN and TF.
