hu.MAP3: Data integration of 25k mass spectrometry experiments accurately identifies protein complexes

Macromolecular protein complexes carry out most functions in the cell including essential functions required for cell survival. The identification of protein complexes in human cells is a long-standing challenge. To address this challenge several high throughput techniques have been developed including affinity purification mass spectrometry (AP-MS), co-fractionation mass spectrometry (CF-MS), proximity labeling strategies, among others, aimed at identifying protein interactions across the human proteome. As a result of these advances, there exists large scale datasets (>25,000 mass spectrometry experiments) using these methods that cover different cell types, tissues, and parts of the proteome. This presented an opportunity to integrate these data to build a more complete and accurate human protein complex map, which we call hu.MAP3 (Fig1A). We identify over 15,000 protein complexes covering ~3/4ths of the human proteome, a substantial increase over previous complex maps. Of note, we place >1,200 highly uncharacterized proteins into complexes providing testable hypotheses as to their function. Further, we refine our complexes by integrating proteomics co-variation data (ProteomeHD) as well as AlphaFold structural models to identify mutually exclusive protein interactions (Fig1B). We expect our resource to be valuable to the broader research community.