Vít Škrhák

This paper introduces a framework for comparing embedding spaces in molecular biology, demonstrating its utility in pre-filtering machine learning tasks and identifying differential patterns across embeddings.

Seq2Pocket improves binding site prediction by combining pLM fine-tuning, data enhancement, and structure-aware post-processing, achieving state-of-the-art performance on benchmark datasets.

This study uses explainable AI to analyze the energy landscape of protein kinases, revealing that allosteric sites are in neutrally frustrated regions, explaining their detection challenges.

This paper compares sequence-based and structure-based methods for predicting allosteric binding sites in protein kinases, finding that sequence-based models perform poorly on allosteric sites despite good performance on orthosteric ones.

CryptoBench introduces a dataset and benchmark for cryptic protein-ligand binding sites, showing sequence-based methods outperform structure-based ones in prediction accuracy.