Fig. 5: KRASG12D generated molecules over four outer AL cycles for the known and unknown generative processes.
From: Optimizing drug design by merging generative AI with a physics-based active learning framework
A Cumulative plots of the known and unknown specific set enrichment over their inner/outer AL cycles. Vertical dotted lines mark the end of each outer AL cycle, dividing the plot into separate outer cycles. Red dots represent inner AL cycles using the KRAS known or unknown specific sets. B UMAP illustrating the generated molecules over the outer AL cycles for both generation processes. Marker’s styles and colors of molecules from different outer AL cycles and generative processes are indicated in the legend. The two distinct generative processes, are labeled as “known” and “unknown”, the initial-specific set is labeled as “initial” and the multiple outer AL cycles as “outerX”. C Glide gscore histograms of docked molecules across the outer AL cycles of the process initiated with the unknown specific set compared to the Glide gscore histogram of the unknown specific set of KRASG12D. D. Zoomed-in view of the right tail of the histograms in panel C.
