Multi-task learning uncovers robust translation cis-regulatory features

To validate the sequence motifs identified by our multi-task learning model MTtrans, a new 5' UTR library with around 8,000 synthetic 5'UTRs was built to express EGFP. The reads count was used as a proxy of translation rate here to validate the estimated regulatory effect of motifs that we inferred from multiple datasets, proving the robustness of the sequence motifs. 5' UTR-driven-EGFP expressing HEK293T cells were sorted by isolated by Fluorescence-activated cell sorting (FACS) according to the intensity of FITC signal and analyzed using PacBio SMRT sequencing.