AutoShim NO WARRANTY DISCLAIMER Authors David Sullivan (Novartis) Eric Martin (Novartis) Description AutoShim is a collection of programs/scripts for training docking rescoring functions by PLS regression on activity data. Descriptors include the docking score(s), "raw" pharmacophore features, as well as multi-feature descriptors learned using decision trees. Docked-poses for training are selected by a trained scoring function in an iterative model-building/pose-selection process. This distribution contains Unix scripts for driving docking (by DockIt), minimizing (by QXP's Flo+), and pharmacophore-feature extraction (by Magnet). Tools are included for designing and writing the SEA files read by Magnet. Predictive modeling procedures are written in R. Calculations using a public activity dataset and PDB protein structure illustrate the entire docking-to-predictions process. Documentation Included in download Requirement(s) Dockit and Magnet Download(s) autoshim.tar.gz Revision History ---