Transformer-Based Promoter Region Prediction

A transformer encoder reads 300 bp of human DNA and predicts whether it is an RNA polymerase II promoter with its transcription start site (TSS) at a fixed position. Longer sequences are scanned on both strands. Promoters come from the Eukaryotic Promoter Database (EPDnew) and the genome sequence from NCBI (GRCh38). The encoder is first pretrained as a masked language model on unlabelled genome windows. Every result shows the predicted TSS, core promoter motifs, the transformer's attention and a second opinion from a hand-crafted-feature model. A complete MATLAB R2026a version covers feature extraction and evaluation.

Category: Bioinformatics - Genomics and Sequence Analysis (Deep Learning, Transformer, MATLAB, Python)

Bioinformatics Project B02 from ProTechSC Academy, with graphical abstract, video demonstration and project guidance.