OpencanSARchem: chemical registration and standardization pipeline for FAIR integration of bioassay data

Abstract There is an explosion of available chemical and bioactivity information that is utilized by AI and machine learning for drug discovery. For harmonization and registration of disparate sources of these chemical structural data, community resources exist to integrate the available bioactivity information but have limitations. We previously developed canSARchem , a robust solution to meet these needs, but it fell short in its commercial license and computational resource requirements. Here we describe OpencanSARchem , an open-source and computationally efficient standardization and registration pipeline suitable for FAIR (Findable, Accessible, Interoperable, and Reusable) data science. We describe the pipeline and demonstrate that it addresses the limitations of our original canSARchem , while still generating chemically valid tautomers and decreasing computational time 100-fold. We carefully assess tautomeric representations using Gibbs free energy calculations and demonstrate limited practical impact on quality of results in return for extensive speed up and open-accessibility. We describe and assess this as an acceptable compromise to democratize this method and meet the needs of the broader community. We have deployed this method within canSAR.ai, unifying disparate sources of biochemical data to make them freely available to the drug discovery community. Scientific contribution We describe OpencanSARchem, an efficient and open-source standardization and registration pipeline for the integration of diverse structural data. OpencanSARchem addresses the challenge of tautomerism while increasing computational efficiency (100-fold) and remaining independent from commercial software. Our pipeline offers a democratized and FAIR method for harmonization of disparate chemical structural data.