Publications

Recent publications

In our new publication titled "external pageCombining IC₅₀ or K_i Values from Different Sources Is a Source of Significant Noise", which was published recently in J. Chem. Inf. Model., we quantify how large the noise is when combining data from different assays, even when applying maximal curation.

We develop an genetic algorithm to mimic the time splitting approach from in-house data with public data in "external pageSIMPD: an Algorithm for Generating Simulated Time Splits for Validating Machine Learning Approaches", which was published recently in J. Cheminf.

The combination of machine learning and classical force fields can lead to a description of the physical interactions between particles with high accuracy as shown in "external pageHybrid Classical/Machine-Learning Force Fields for the Accurate Description of Molecular Condensed-Phase Systems", which was published recently in Chem. Sci.

We have developed a data-driven approach to assign atomic partial charges extracted with MBIS at DFT level of theory as described in "external pageDASH: Dynamic Attention-Based Substructure Hierarchy for Partial Charge Assignment", which was published recently in J. Chem. Inf. Model.