Full workings here: https://github.com/QuantumCorona/AutodockVina_3CL-pro/tree/master/0009_hpc_fragment_outputs
These structures were not looked at / post-processed by a Chemist, so some discussion & redesign using human intelligence might be really useful! (I trained as a physicist, so don’t really have a gut-feeling for these things.)
We ran the highest scoring 200 molecules, downloaded from the github link above, through our predictive model; this is the same model used in conjunction with the JAR-KUA-672 submission. Overall these look very promising, as is illustrated in the histogram below. For comparison note that on 1M random ENAMINE real molecules our model identified only around 200 hits, i.e. molecules scoring > 0.5. 