Fig. 3: Interpretation of the gene expression-predictive models shows important predictive nucleotides are located mainly in the regions near the TSS and TTS across different crop plants for the leaf SSR models.
From: Deep learning the cis-regulatory code for gene expression in selected model plants
a Average per-base importance scores computed with DeepLIFT22 for A. thaliana, S. lycopersicum, S. bicolor and Z. mays SSR leaf models. Importance scores were averaged across validation sequences depicting the average importance of regions with the 3000 nt long sequences. b Per-base importance scores were used for motif generation with TF-MoDISco23. The resulting expression-predictive motifs (EPMs) share consensus sequences following the extended IUPAC nucleotide code (Supplementary Data 4) and position-specific occurrences around the TSS and TTS region.
