The Broad Institute's Connectivity Map is a collection of genome-wide, transcriptional expression data from cultured human cells. The data is used to find the functional connections between drugs, genes, and diseases through the transitory feature of common gene-expression changes.
The goal of this 2016 precision medicine contest on Topcoder was to maximize the accuracy of the inferred gene expression values while minimizing the number of measured gene expressions. The results will further expand horizons for computational biologists and scientists who seek to find drugs that cure diseases.
In this challenge, 88 participants competed for $20,000 and submitted 1,116 solutions (see the problem statement). The results from the competitors showed at least modest improvement in accuracy, but some for some genes, accuracy improved dramatically.
Contest winners generated solutions that were combined into ensembles ranging from k-nearest neighbour models to neural networks. The top five produced between 60-63% of the gene-level predictions better than the average of the landmark genes with the ensemble solution predicting closer to 70%. Five of the top 10 winners had no experience in biology or computational biology.