Linear Programming and Machine Learning Benchmarks

This page is meant to provide access to various Linear Programming and Machine Learning problems and benchmarks that we generated mostly in the context of protein structure prediction. The problems included below may be downloaded to train (optimize) your own classifiers or scoring functions for protein structure predictions. We intend to update this page frequently, reporting the best results achieved for each problem using LP and ML techniques. In order to download a specific problem please follow the links below:

The Rosetta decoy set recognition problem with standard pairwise contact potential  >>>.

Protein secondary structure prediction problem, using linear combination of weak, statistically derived classifiers >>>.

Trans-membrane domain prediction problem, using a linear combination of weak, statistically derived classifiers >>>.

More to come ...

Most of the benchmarks included here were  generated using our LOOPP program for the design of scoring functions for protein folding and protein structure prediction. The relevant papers and references include (see Publications page for our group):

R. Adamczak and J. Meller; On the Transferability of Folding and Threading Potentials and Sequence-Independent Filters for Protein Folding Simulations, submitted

A. Porollo, R. Adamczak, M. Wagner and J. Meller; Maximum Feasibility Approach for Consensus Classifiers: Applications to Protein Structure Prediction, submitted

M. Wagner, J. Meller and R. Elber; Large-Scale Linear Programming Techniques for the Design of Protein Folding Potentials, Mathematical Programming, to appear (2003)

J. Meller, M. Wagner and R. Elber, Maximum Feasibility Guideline to the Design and Analysis of Protein Folding Potentials, Journal of Computational Chemistry 23: 111-118 (2002) 

J. Meller and R. Elber, Linear Programming Optimization and a Double Statistical Filter for Protein Threading Potentials, Proteins Struct. Fun. Gen. 45: 241-261 (2001) 

J. Meller and R.Elber; LOOPP - Lerning, Observing and Outputting Protein Patterns, a public domain package for protein recognition and designing folding potentials, available as part of NIH resources at Cornell Theory Center, Ithaca 2000 


Author: Jarek Meller