Unbiased Protein Association Study on the Public Human Proteome Reveals Biological Connections between Co-Occurring Protein Pairs

Mass-spectrometry-based, high-throughput
proteomics experiments
produce large amounts of data. While typically acquired to answer
specific biological questions, these data can also be reused in orthogonal
ways to reveal new biological knowledge. We here present a novel method
for such orthogonal data reuse of public proteomics data. Our method
elucidates biological relationships between proteins based on the
co-occurrence of these proteins across human experiments in the PRIDE
database. The majority of the significantly co-occurring protein pairs
that were detected by our method have been successfully mapped to
existing biological knowledge. The validity of our novel method is
substantiated by the extremely few pairs that can be mapped to existing
knowledge based on random associations between the same set of proteins.
Moreover, using literature searches and the STRING database, we were
able to derive meaningful biological associations for unannotated
protein pairs that were detected using our method, further illustrating
that as-yet unknown associations present highly interesting targets
for follow-up analysis.