In this presentation I will discuss a case study illustrating how we use machine learning [1] and data visualization methods such as TMAP [2] and Faerun [3] to guide the discovery of new bioactive compounds. I will focus on a project aimed at discovering membrane disruptive, non-hemolytic anticancer peptides by using a genetic algorithm and recurrent neural networks (RNNs). The RNNs were trained with experimental data from the DBAASP (Database of Antimicrobial Activity and Structure of Peptides) [4] and used for both sequence generation and hemolysis/activity classification. The approach helped at identifying peptides acting selectively on cancer cells among many amphiphilic, α-helical, and generally membrane disruptive sequences. 1. A. Capecchi, X. Cai, H. Personne, T. Köhler, C. van Delden and J.-L. Reymond, Chem. Sci., 2021, 12, 9221–9232. 2. D. Probst and J.-L. Reymond, J. Cheminf., 2020, 12, 12. 3. D. Probst, J.-L. Reymond and J. Wren, Bioinformatics, 2018, 34, 1433–1435. 4. G. Gogoladze, M. Grigolava, B. Vishnepolsky, M. Chubinidze, P. Duroux, M.-P. Lefranc and M. Pirtskhalava, FEMS Microbiol Lett, 2014, 357, 63–68.

Download the slides for this talk.Download ( PDF, 4596.5 MB)