Discovery and Basic Research
Alexander Tropsha, PhD
KH Lee Distinguished Professor
University of North Carolina
Chapel Hill, North Carolina
Recent growth of the experimental chemical bioactivity datasets has enabled meaningful and expanding application of AI methods to support pharmaceutical research. Along with exciting applications, there have also beеn concerns about the value and predictive power of AI methods as applied to the analysis of big data especially with respect to how data quality may affect model accuracy. While the problem of rigor and reproducibility of biomedical research has been acknowledged by the NIH, there have been assertions in the research literature and associated social media that AI methods can overcome the problem of “dirty” data. I shall discuss the importance of data curation for the development of reliable models and address workflows for chemical and biological data curation and validated computational model development. I will also discuss the need to establish criteria for rigor and reproducibility of computational models that could support their reliable applications both as discovery as well as regulatory support tools.