Florent Laval

Genotype-phenotype relationships are mediated through intricate networks of physical and functional interactions among macromolecules. Knowledge of the interactome is vital to understand and model genetics and cellular biology. Recent advances in accurately predicting tertiary protein structures using artificial intelligence (AI) approaches such as AlphaFold ¹ have revived the vision that the protein-protein interactome might be fully predictable through computational modeling of quaternary structures. Here we present a comprehensive experimental framework to systematically assess the impact of AI-driven interactome predictions for yeast ² and human ³ . We find that the quality of high-confidence predictions is on par with established experimental approaches. However, in proteome-wide screening, the tested AI approaches underperform in the discovery of strictly novel protein-protein interactions (PPIs) compared to experimental reference interactome maps. In particular, the yeast interactome map described here identifies >40-fold more novel PPIs than its AI counterpart. Strikingly, AlphaFold provides structural models for a substantial number of experimentally identified PPIs missed by the virtual screens. Our results suggest that, at this stage, the main contribution of AI predictions is to provide quaternary structure models for experimentally identified PPIs.

Over 90% of missense variants across ∼4,000 disease-associated genes are variants of uncertain significance (VUS). Experimental variant effect measurements provide critical evidence about pathogenicity and inform disease biology, but most variants lack data and clinical translation has been limited. The Impact of Genomic Variation on Function Consortium generated experimental data for 62,215 variants across ten genes using multiplexed assays and 1,407 variants across 163 genes using arrayed assays, curated 193,139 additional community-generated variant effect measurements across 30 additional genes, and developed automated calibration methods for translating experimental data and variant effect predictions into clinical evidence. To reduce current VUS, we developed a scalable workflow using only experimental and predictive evidence, enabling reclassification of 75% of the 16,115 VUS in these genes as pathogenic or benign with 90,000 unobserved variants; 62% had enough evidence to be “preclassified” as pathogenic or benign. We validated our data, evidence and classifications using All of Us and created interactive resources to enable clinical use of the calibrated data. Thus, for 40 genes, representing 1% of the clinical genome, we resolve most existing VUS and future variants, illustrating how systematic use of scalable evidence can empower genomic medicine.

Disruption of protein–protein interactions (PPIs) is a major mechanism of a variant’s deleterious effect. Computational tools are needed to assess such variants at scale, yet existing predictors rarely consider loss of specific interactions, particularly when variants perturb binding interfaces without significantly affecting protein stability. To address this problem, we present MutPred-PPI, a graph attention network that predicts interaction-specific (edgetic) effects of missense variants by operating on AlphaFold 3-based protein complex contact graphs with protein language model embeddings imposed upon nodes. We systematically evaluated our model with stringent group cross-validation as well as benchmark data recently collected within the IGVF Consortium. MutPred-PPI outperformed all baseline methods across all evaluation criteria, achieving an AUC of 0.85 on seen proteins and 0.72 on previously unseen proteins in cross-validation, demonstrating strong generalizability despite scarce training data. To demonstrate biomedical relevance, we applied MutPred-PPI to variants from ClinVar, HGMD, COSMIC, gnomAD, and two de novo neurodevelopmental disorder-linked datasets. Disease-associated variants from Clin-Var and HGMD showed strong enrichment for both quasi-null and edgetic effects, whereas population variants from gnomAD increasingly preserved interactions with higher allele frequencies. Notably, we observed a strong edgetic disruption signature in highly recurrent cancer variants from both the full COSMIC dataset and a subset of variants from oncogenes. Recurrent tumor suppressor gene variants and autism spectrum disorder-associated variants exhibited moderate quasi-null enrichment, whilst neurodevelopmental disorder-linked variants showed a weak edgetic disruption signature. These results indicate distinct PPI perturbation mechanisms across disease types and show that MutPred-PPI captures functionally relevant molecular effects of pathogenic variants.

Legionella pneumophila uses over 300 translocated effector proteins to rewire host cells during infection and create a replicative niche for intracellular growth. To date, several studies have identified L. pneumophila effectors that indirectly and directly regulate the activity of other effectors, providing an additional layer of regulatory complexity. Among these are “metaeffectors,” a special class of effectors that regulate the activity of other effectors once inside the host. A defining feature of metaeffectors is direct, physical interaction with a target effector. Metaeffector identification, to date, has depended on phenotypes in heterologous systems and experimental serendipity. Using a multiplexed, recombinant barcode-based yeast two-hybrid technology we screened for protein–protein interactions among all L. pneumophila effectors and 28 components of the Dot/Icm type IV secretion system (>167,000 protein combinations). Of the 52 protein interactions identified by this approach, 44 are novel protein interactions, including 10 novel effector–effector interactions (doubling the number of known effector–effector interactions). IMPORTANCE: Secreted bacterial effector proteins are typically viewed as modulators of host activity, entering the host cytosol to physically interact with and modify the activity of one or more host proteins in support of infection. A growing body of evidence suggests that a subset of effectors primarily function to modify the activities of other effectors inside the host. These “effectors of effectors” or metaeffectors are often identified through experimental serendipity during the study of canonical effector function against the host. We previously performed the first global effector-wide genetic interaction screen for metaeffectors within the arsenal of Legionella pneumophila , an intracellular bacterial pathogen with over 300 effectors. Here, using a high-throughput, scalable methodology, we present the first global interaction network of physical interactions between L. pneumophila effectors. This data set serves as a complementary resource to identify and understand both the scope and nature of non-canonical effector activity within this important human pathogen.

Assessment of the impact of missense variants on protein subcellular localization

Legionella pneumophila uses over 300 translocated effector proteins to rewire host cells during infection and create a replicative niche for intracellular growth. To date, several studies have identified L. pneumophila effectors that indirectly and directly regulate the activity of other effectors, providing an additional layer of regulatory complexity. Amongst these are “metaeffectors” – a special class of effectors that regulate the activity of other effectors once inside the host. A defining feature of metaeffectors is direct, physical interaction with a target effector. Metaeffector identification to date has depended on phenotypes in heterologous systems and experimental serendipity. Using a multiplexed, recombinant-barcode-based yeast two-hybrid technology we screened for protein-protein interactions amongst all L. pneumophila effectors and several components of the Dot/Icm type IV secretion system (>167,000 protein combinations). Of the 52 protein interactions identified by this approach, 44 are novel protein interactions, including ten novel effector-effector interactions (doubling the number of known effector-effector interactions).

Summary: Most human Transcription factors (TFs) genes encode multiple protein isoforms differing in DNA binding domains, effector domains, or other protein regions. The global extent to which this results in functional differences between isoforms remains unknown. Here, we systematically compared 693 isoforms of 246 TF genes, assessing DNA binding, protein binding, transcriptional activation, subcellular localization, and condensate formation. Relative to reference isoforms, two-thirds of alternative TF isoforms exhibit differences in one or more molecular activities, which often could not be predicted from sequence. We observed two primary categories of alternative TF isoforms: “rewirers” and “negative regulators”, both of which were associated with differentiation and cancer. Our results support a model wherein the relative expression levels of, and interactions involving, TF isoforms add an understudied layer of complexity to gene regulatory networks, demonstrating the importance of isoform-aware characterization of TF functions and providing a rich resource for further studies.

Understanding how genetic variation affects phenotypes represents a major challenge, particularly in the context of human disease. Although numerous disease-associated genes have been identified, the clinical significance of most human variants remains unknown. Despite unparalleled advances in genomics, functional assays often lack sufficient throughput, hindering efficient variant functionalization. There is a critical need for the development of more potent, high-throughput methods for characterizing human genetic variants. Here, we review how yeast helps tackle this challenge, both as a valuable model organism and as an experimental tool for investigating the molecular basis of phenotypic perturbation upon genetic variation. In systems biology, yeast has played a pivotal role as a highly scalable platform which has allowed us to gain extensive genetic and molecular knowledge, including the construction of comprehensive interactome maps at the proteome scale for various organisms. By leveraging interactome networks, one can view biology from a systems perspective, unravel the molecular mechanisms underlying genetic diseases, and identify therapeutic targets. The use of yeast to assess the molecular impacts of genetic variants, including those associated with viral interactions, cancer, and rare and complex diseases, has the potential to bridge the gap between genotype and phenotype, opening the door for precision medicine approaches and therapeutic development.

Comprehensive understanding of the human protein-protein interaction (PPI) network, aka the human interactome, can provide important insights into the molecular mechanisms of complex biological processes and diseases. Despite the remarkable experimental efforts undertaken to date to determine the structure of the human interactome, many PPIs remain unmapped. Computational approaches, especially network-based methods, can facilitate the identification of previously uncharacterized PPIs. Many such methods have been proposed. Yet, a systematic evaluation of existing network-based methods in predicting PPIs is still lacking. Here, we report community efforts initiated by the International Network Medicine Consortium to benchmark the ability of 26 representative network-based methods to predict PPIs across six different interactomes of four different organisms: A. thaliana, C. elegans, S. cerevisiae, and H. sapiens. Through extensive computational and experimental validations, we found that advanced similarity-based methods, which leverage the underlying network characteristics of PPIs, show superior performance over other general link prediction methods in the interactomes we considered.

Understanding the mechanisms of coronavirus disease 2019 (COVID-19) disease severity to efficiently design therapies for emerging virus variants remains an urgent challenge of the ongoing pandemic. Infection and immune reactions are mediated by direct contacts between viral molecules and the host proteome, and the vast majority of these virus–host contacts (the ‘contactome’) have not been identified. Here, we present a systematic contactome map of severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) with the human host encompassing more than 200 binary virus–host and intraviral protein–protein interactions. We find that host proteins genetically associated with comorbidities of severe illness and long COVID are enriched in SARS-CoV-2 targeted network communities. Evaluating contactome-derived hypotheses, we demonstrate that viral NSP14 activates nuclear factor κB (NF-κB)-dependent transcription, even in the presence of cytokine signaling. Moreover, for several tested host proteins, genetic knock-down substantially reduces viral replication. Additionally, we show for USP25 that this effect is phenocopied by the small-molecule inhibitor AZ1. Our results connect viral proteins to human genetic architecture for COVID-19 severity and offer potential therapeutic targets.