Kevin Drew

Protein complexes carry out critical functions during development. Mapping cell-type-specific protein complexes in early embryogenesis is essential for uncovering the molecular mechanisms of embryonic development. We unfortunately currently lack a catalog of protein complexes specific for embryogenesis including embryonic stem cells. Here we develop a proteomics-based machine learning workflow to predict protein complexes present in mouse embryonic stem cells (mESC). Our method integrates data from a cell-type agnostic protein complex resource with co-fractionation mass spectrometry (CF-MS) experiments on mESC. Our resource, stem.MAP, identifies 3,986 complexes encompassing ∼133,000 unique interactions, and 8,616 proteins. stem.MAP captures variant complexes critical for stem cell function including the PRC2, BAF and NuRD complexes. stem.MAP also discriminates mutually exclusive interactions specific to embryonic stem cells (e.g. Eif4ebp2 - Eif4e). Finally we observe 1,388 high confidence stem.MAP interactions found only in mESC CF-MS data, many with support from AlphaFold modeling and a gastruloid coregulation network. We make stem.MAP protein complexes available through our web resource: stem.proteincomplexes.org. We anticipate stem.MAP will be beneficial to the study of early development and we expect our proteomics-based machine learning workflow to be applicable to the creation of any context-specific protein interaction network.

Cellular function is driven by the activity of proteins in stable complexes. Protein complex assembly depends on the direct physical association of component proteins. Advances in macromolecular structure prediction with tools like AlphaFold and RoseTTAFold have greatly improved our ability to model these interactions in silico, but an all-by-all analysis of the human proteome’s ~200 M possible pairs remains computationally intractable. A comprehensive cellular map of direct protein interactions will therefore be an invaluable resource to direct screening efforts. Here, we present DirectContacts2 , a machine learning model that distinguishes direct from indirect protein interactions using features derived from over 25,000 mass spectrometry experiments. Applied to ~25 million human protein pairs, our model outperforms previous resources in identifying direct physical interactions and enriches for accurate structural models including ~2500 AlphaFold3 models. Our framework enables structural modeling of disease-relevant complexes (e.g. orofacial digital syndrome (OFDS) complex) offering insights into the molecular consequences of pathogenic mutations (OFD1) and broadly, establishes a highly accurate protein wiring diagram of the cell.

An outstanding challenge in molecular biology is the production of a complete and accurate set of protein interactions for a given organism. Todor et al, introduce a clever strategy to efficiently screen all protein interactions in Mycoplasma genitalium using AlphaFold3 (Todor et al, 2026). K. Drew highlights a recent study by Todor and colleagues (in this issue of Molecular Systems Biology) that presents an innovative strategy to systematically map all protein interactions in M. genitalium using AlphaFold3.

Cellular function is driven by the activity proteins in stable complexes. Protein complex assembly depends on the direct physical association of component proteins. Advances in macromolecular structure prediction with tools like AlphaFold and RoseTTAFold have greatly improved our ability to model these interactions in silico, but an all-by-all analysis of the human proteome’s ~200M possible pairs remains computationally intractable. A comprehensive cellular map of direct protein interactions will therefore be an invaluable resource to direct screening efforts. Here, we present DirectContacts2 , a machine learning model that distinguishes direct from indirect protein interactions using features derived from over 25,000 mass spectrometry experiments. Applied to ~26 million human protein pairs, our model outperforms previous resources in identifying direct physical interactions and enriches for accurate structural models including ~2,500 new AlphaFold3 models. Our framework enables structural modeling of disease-relevant complexes (e.g. orofacial digital syndrome (OFDS) complex) offering insights into the molecular consequences of pathogenic mutations (OFD1) and broadly, establishes a highly accurate protein wiring diagram of the cell.

Cilia are essential organelles, and variants in genes governing ciliary function result in ciliopathic diseases. The Ciliogenesis and PLANar polarity Effectors (CPLANE) protein complex is essential for ciliogenesis, and all but one subunit of the CPLANE complex have been implicated in human ciliopathy. Here, we identify three families in which variants in the remaining CPLANE subunit CPLANE2/RSG1 also cause ciliopathy. These patients display cleft palate, tongue lobulations and polydactyly, phenotypes characteristic of Oral-Facial-Digital Syndrome. We further show that these alleles disrupt two vital steps of ciliogenesis, basal body docking and recruitment of intraflagellar transport proteins. Moreover, APMS reveals that Rsg1 binds CPLANE and the transition zone protein Fam92 in a GTP-dependent manner. Finally, we show that CPLANE is generally required for normal transition zone architecture. Our work demonstrates that CPLANE2/RSG1 is a causative gene for human ciliopathy and also sheds new light on the mechanisms of ciliary transition zone assembly.

Macromolecular protein complexes carry out most cellular functions. Unfortunately, we lack the subunit composition for many human protein complexes. To address this gap we integrated >25,000 mass spectrometry experiments using a machine learning approach to identify >15,000 human protein complexes. We show our map of protein complexes is highly accurate and more comprehensive than previous maps, placing nearly 70% of human proteins into their physical contexts. We globally characterize our complexes using mass spectrometry based protein covariation data (ProteomeHD.2) and identify covarying complexes suggesting common functional associations. hu.MAP3.0 generates testable functional hypotheses for 472 uncharacterized proteins which we support using AlphaFold modeling. Additionally, we use AlphaFold modeling to identify 5871 mutually exclusive proteins in hu.MAP3.0 complexes suggesting complexes serve different functional roles depending on their subunit composition. We identify expression as the primary way cells and organisms relieve the conflict of mutually exclusive subunits. Finally, we import our complexes to EMBL-EBI’s Complex Portal ( https://www.ebi.ac.uk/complexportal/home ) and provide complexes through our hu.MAP3.0 web interface ( https://humap3.proteincomplexes.org/ ). We expect our resource to be highly impactful to the broader research community.

Macromolecular protein complexes carry out most functions in the cell including essential functions required for cell survival. Unfortunately, we lack the subunit composition for all human protein complexes. To address this gap we integrated >25,000 mass spectrometry experiments using a machine learning approach to identify > 15,000 human protein complexes. We show our map of protein complexes is highly accurate and more comprehensive than previous maps, placing ∼75% of human proteins into their physical contexts. We globally characterize our complexes using protein co-variation data (ProteomeHD.2) and identify co-varying complexes suggesting common functional associations. Our map also generates testable functional hypotheses for 472 uncharacterized proteins which we support using AlphaFold modeling. Additionally, we use AlphaFold modeling to identify 511 mutually exclusive protein pairs in hu.MAP3.0 complexes suggesting complexes serve different functional roles depending on their subunit composition. We identify expression as the primary way cells and organisms relieve the conflict of mutually exclusive subunits. Finally, we import our complexes to EMBL-EBI’s Complex Portal ( https://www.ebi.ac.uk/complexportal/home ) as well as provide complexes through our hu.MAP3.0 web interface ( https://humap3.proteincomplexes.org/ ). We expect our resource to be highly impactful to the broader research community.

Mutations in retinal primary cilia are responsible for human blindness but the mechanisms are not fully understood (Wheway et al., 2014). Characterizing the proteome of an organelle such as cilia, is a fruitful way to understand its function but methods often require considerable sample quantities. Here we develop a method to isolate the primary cilia of photoreceptor cells from bovine retinas. Through LC/MS/MS proteomics analysis we identify proteins enriched for cilia function including ciliopathy disease. This study shows our method can be used to isolate retinal primary cilia to obtain sufficient quantities of native protein samples.