Sera Aylin Cakiroglu

The paper introduces a contrastive fine-tuning method for single-cell foundation models to detect subtle transcriptomic changes in Perturb-seq data, outperforming traditional differential expression methods.

The paper introduces dynamic token adaptation to combine single-cell foundation models with DNA-sequence embeddings to predict gene regulation in different genetic contexts.

ChromWave predicts TF and nucleosome occupancy competition using deep learning, showing accurate predictions in yeast and human promoters.