Download
Implementation

A sample implementation of the MEFIT system is provided here for research purposes. If you find MEFIT useful, let us know! For that matter, if you find bugs, have questions, or are interested in source code for MEFIT, let us know that, too. We can be contacted at chuttenh@princeton.edu and ogt@princeton.edu. We are currently able to offer for download:

Predictions

Each of the following files contains MEFIT's pairwise functional relationship predictions for one biological function. Files are formatted to contain one gene pair and probability of relationship per line. Two prediction sets are available for each biological process, one generated from our test set (with the addition of genes annotated to GO's "biological process unknown") and one generated over the entire yeast genome.

Functions

Gold Standard

This file represents the gold standard answer file generated from our 200 biological functions. Each line contains a gene pair and a 0 (for unrelated pairs) or a 1 (related pairs). As described in Huttenhower et al. 2006, gene pairs coannotated to any function were considered to be related, and gene pairs not coannotated within a very general set of functions were considered to be unrelated. All other gene pairs were excluded from the answer file.

Data

The MEFIT predictions displayed here (and discussed in our paper) used the following microarray data sets as input.