Fig. 6. Analysis of all oligomers predicted here across all runs. (a) Examples of the best performing oligomers with their values of ES1, IP and fosc,S1. (b) The best and worst performing fragments based on the average Fcomb of molecules containing that fragment. (c) Distribution of the combined property grouped by oligomers with building blocks from different clusters. Here we define the dominant building block cluster for a molecule as the cluster with the highest number of building block present in the molecule. If no cluster is dominant, we label the molecule as “mixed.” The grey violin plots show the data in the benchmark dataset. The lines divide the distribution into quartiles. (d) number of molecules in the dataset grouped by dominant building block cluster. The grey/darker bins show the data in the benchmark dataset.
