Overall, SRC-VLS outperformed the "random picking" baseline (as expected); however it was unable to guarantee a systematic ranking of true binders among the first hits.
All 5 metrics rely on the so-called accumulation curve, Fa k), where the subscript "a" stands for active molecule, which represents the count, possibly normalized, of how many true binders obtained a rank better or equal to a given one in a docking run; Fa(k) contains all the information needed to assess the performance of a run, and a combination of multiple runs as well.
This is particularly relevant for VLS protocols, where early recognition of true binders is a major achievement of this computational drug discovery approach.
Three groups evince the behavior expected for true binders to a single antibody majority of sequences occur in categories corresponding to a single antibody, and relative cluster sizes are significantly increased in selection rounds.
This highlighted that the newly adopted set-ups enable robust discrimination between true binders and nonbinders, which is a critical requisite in biophysical fragment screening.
FINDSITEcomb will be helpful in enriching true binders for 65% of the targets in a typical genome sequence.
As expected, the analysis of the results confirmed that true binders were ranked at the first positions, when a suitable receptor conformation was used.
For our large scale analysis, we have chosen a relatively high combined score threshold of 1.25, which gives preferentiality to true binders.
For example, in some cases, a sequence evincing the behavior expected for phages with exceptional ability to multiply is contained within a cluster the overall category profile of which falls into the category of true binders.
Conversely, SRC-VLS could not identify true binders when non-cognate receptor conformations (or similar) were used.
