Here are some cut-and-paste transcripts from the developers’ email archives:
“In the original reference for Maxsub (Siew et al, 2000, http://bioinformatics.oxfordjournals.org/cgi/reprint/16/9/776.pdf), the programs takes a single “RMSD threshold” input parameter (default 3.5A) and tries various superimpositions to find the largest number of CA atoms in the model that can be fit to the native under this threshold. After the optimal superimposition is found, a normalized score is produced to reflect the quality of the model with 1.0 as perfect prediction and 0.0 as completely incorrect. “
…
“our maxsub is charlie’s implementation (that he did in mammoth), and is not exactly the same as danny fischer’s. not sure what the difference between maxsub and maxsub2.0 are.”
So – Maxsub is similar to whatever’s in that reference, probably identical in spirit but perhaps slightly different in implementation.