Long-range information in RNA embeddings

#1
by Shicheng - opened

Training a 14.1M-parameter Mamba model on 1.96 million RNAcentral sequences with an 8,192-nucleotide context creates an interesting efficient baseline for long RNA embeddings. Have you evaluated whether the representations retain long-range structural or regulatory information well enough for ncRNA function prediction or sequence-level therapeutic target assessment?

Sign up or log in to comment