Dannie Durand

Motivation: Biomolecular sequences can be represented as strings over an alphabet, an analogy that has motivated many applications of computational linguistic techniques to biological problems. However, such methods must be adapted to the characteristic scale and organization of biomolecular data. Here, we consider the problem of bigram smoothing for multidomain protein architectures, where domain bigram frequency data is extremely sparse and differs from textual data in alphabet size, string length distribution, the relationship between bigram and unigram frequencies, tandem repeat lengths, and the distribution of domain adjacencies. Moreover, some domain combinations are unobserved because they are biologically incompatible, others because the data are incomplete. A smoothing method that distinguishes these two cases is required. Results: We propose a unified smoothing framework based on interpolation that can be tuned to accommodate different bigram data characteristics. Within this framework, we design specific model variants suited to protein domain bigram data: these assign low adjusted counts to pairs that are likely incompatible, while making appropriate adjustments for undersampled pairs. We demonstrate empirically that this approach distinguishes the two cases while preserving the characteristic signatures of multidomain data. Availability and implementation: Implementations of smoothing methods, the scripts used to generate all results presented in this paper, and the curated lists of extracellular and DNA-binding domains are available at https://codeberg.org/xcui297/protein-domain-smoothing .

Background: The Critical Assessment of Functional Annotation (CAFA) is a community effort held to understand the field of computational protein function prediction. Every three years, since 2010, the organizers initiate an experiment to collect function predictions on a large set of proteins and then evaluate the performance of predicting methods on a subset of proteins that have accumulated experimental annotations between the submission deadline and the evaluation time. CAFA provides an independent and rigorous assessment of the current state of the art, thus leveling the playing field, highlighting successes, revealing bottlenecks, and offering a forum for the exchange of ideas in protein science. Here, we report the results of the fourth CAFA experiment (CAFA4). Results: CAFA4 featured the participation of 148 methods from 70 research groups on a total of 46,205 unique proteins over a 5-year annotation accumulation phase, the longest in any CAFA. In a comparison across CAFA2-CAFA4 methods, the prediction of Gene Ontology (GO) terms has clearly improved across all three GO aspects and traditional evaluation settings. While not achieving the first rank, several CAFA2 and CAFA3 methods featured in the top ten methods in many evaluations, suggesting that earlier methods still hold relevance. The performance is weaker in the newly introduced “partial knowledge” evaluation category (proteins with experimental annotations before submission deadline that gained additional annotations in the same GO aspect during the annotation accumulation phase), highlighting the need for a new class of methods. The rankings of the methods were stable over the years in traditional evaluation settings, but less so in the new partial knowledge evaluation. Overall, the field continues to progress with some influx of new participants. Sustained efforts will be necessary to substantially advance it.

The Critical Assessment of Functional Annotation (CAFA) is a long-standing community effort to independently assess computational methods for protein function prediction, to highlight well-performing methodologies, to identify bottlenecks in the field, and to provide a forum for the dissemination of results and exchange of ideas. In its fifth round (CAFA5) of triennial challenges, a partnership with Kaggle Inc. facilitated participation from a large community of data scientists and computational biologists through a competitive prospective challenge on the crowdsourcing platform. In this work, we present an in-depth analysis of the submitted predictions and report improvements in accuracy over all methods from the previous CAFA challenges. We further introduce a new evaluation setting for proteins with pre-existing (incomplete) annotations and identify the need for methods that better leverage existing annotations to predict those that will be discovered later. Finally, we characterize the prospective evaluation framework by examining performance on a strict set of unpublished annotations and across intermediate database releases. Our results indicate that recent developments in the field, such as the availability of protein language models and accurately predicted 3D structures, as well as the growth of experimental annotations through biocuration, have all contributed to performance improvements. ¹

Multidomain proteins are mosaics of domains, protein modules that are associated with a specific structure or function and are found in diverse combinations. This modular organization facilitates the evolution of novel protein functions, but the principles that govern the relationship between the domain content of a protein and its function is poorly understood. In particular, do domains always contribute the same function, or does the functional contribution of a domain depend on the neighboring domains in the protein? To answer this question, we used vector embeddings, which account for local contextual signals, to model the protein domain content of multidomain proteins. We observe that multidomain architectures that are semantically similar share more functional attributes than multidomain architectures selected based on domain content similarity, alone, suggesting that context is important for understanding the relationship between domain content and protein function. Surprisingly, vector semantics also identified multidomain architecture pairs with significantly high functional similarity, despite having no domains in common at all, suggesting that vector semantics may be discovering domain “synonyms”. Taken together, our results underscore the importance of contextual models for understanding the interplay between domain architecture evolution and functional innovation in multidomain proteins.

Reconstruction of the ancestral protein repertoire offers valuable insights into the tempo and mode of protein content evolution, but can be highly sensitive to model choice. We used a phylogenetic Birth-Death-Gain model to investigate the evolution of the metazoan protein domain repertoire. Domains, protein modules with a distinct structure and function, represent the basic components of protein repertoire. Given a species tree and a census of protein domain families in present-day species, we estimated the most likely rates of domain family origination, duplication, and loss. Rates were allowed to vary across species lineages and domain families, decoupling these factors. Statistical hierarchical clustering of family-specific rates reveals groups of domains evolving in concert. Moreover, we observe a strong and significant association between family rate and family function. Interestingly, families with functions associated with metazoan innovations tend to have the fastest rates. We further inferred the expected ancestral domain content and the history of domain family gains, losses, expansions, and contractions in each species lineage. Our analysis reveals an ongoing process of domain family replacement and resizing, consistent with extensive remodeling of the protein domain repertoire. This stands in contrast to recent reports of widespread loss during metazoan evolution, which were obtained with more constrained models. The use of a powerful, probabilistic Birth-Death-Gain model reveals an unexpected level of genomic plasticity.

Multidomain proteins are mosaics of domains , protein modules that are associated with a specific structure or function and are found in diverse combinations. This modular organization facilitates the evolution of novel protein functions, but the principles that govern the relationship between the domain content of a protein and its function is poorly understood. In particular, do domains always contribute the same function, or does the functional contribution of a domain depend on the neighboring domains in the protein? To answer this question, we used vector embeddings, which account for local contextual signals, to model the protein domain content of multidomain proteins. We observe that multidomain architectures that are semantically similar share more functional attributes than multidomain architectures selected based on domain content similarity, alone, suggesting that context is important for understanding the relationship between domain content and protein function. Surprisingly, vector semantics also identified multidomain architecture pairs with significantly high functional similarity, despite having no domains in common at all, suggesting that vector semantics may be discovering domain “synonyms”. Taken together, our results underscore the importance of contextual models for understanding the interplay between domain architecture evolution and functional innovation in multidomain proteins.

Domains, sequence fragments that encode protein modules with a distinct structure and function, are the basic building blocks of proteins. The set of domains encoded in the genome serves as the functional toolkit of the species. Here, we use a phylogenetic Birth-Death-Gain model to investigate the evolution of this protein toolkit in metazoa. Given a species tree and the set of protein domain families in each present-day species, this approach estimates the most likely rates of domain origination, duplication and loss. Statistical hierarchical clustering of domain family rates reveals sets of domains with similar rate profiles, consistent with groups of domains evolving in concert. Moreover, we find that domains with similar functions tend to have similar rate profiles. Interestingly, domains with functions associated with metazoan innovations, including immune response, cell adhesion, tissue repair, and signal transduction, tend to have the fastest rates. We further infer the expected ancestral domain content and the history of domain family gains, losses, expansions, and contractions on each branch of the species tree. In contrast to recent reports of widespread loss during metazoan evolution, we observe little evidence of genome streamlining. Rather, our analysis reveals an ongoing process of domain family replacement and resizing, consistent with extensive remodeling of the protein domain repertoire. The use of a powerful, probabilistic Birth-Death-Gain model reveals an unexpected level of genomic plasticity and a striking harmony between the evolution of domain usage in metazoan proteins and organismal innovation.

SUMMARY: All eukaryotes share a single-celled ancestor from ∼1.5–1.8 billion years ago, the Last Eukaryotic Common Ancestor (LECA). Roughly half of gene families found in modern eukaryotes were already present in LECA, forming molecular systems that continue to influence genetic diseases and traits today. To investigate these systems, we compared genes across 156 organisms to define a core set of protein-coding gene families likely present in LECA, with a quarter remaining uncharacterized. Integrating >26,000 mass spectrometry proteomics analyses from 31 species, we inferred higher-order complexes among these ancient proteins. This reconstructed interactome reveals both established and novel assemblies, offering a biochemical snapshot of LECA’s organization. Finally, by exploring these ancient protein interactions, we found new human gene-disease associations for bone density and congenital birth defects, illustrating the value of ancestral protein networks for modern functional genetics.

Homotypic membrane fusion catalyzed by the atlastin (ATL) GTPase sustains the branched endoplasmic reticulum (ER) network in metazoans. Our recent discovery that two of the three human ATL paralogs (ATL1/2) are C-terminally autoinhibited implied that relief of autoinhibition would be integral to the ATL fusion mechanism. An alternative hypothesis is that the third paralog ATL3 promotes constitutive ER fusion with relief of ATL1/2 autoinhibition used conditionally. However, published studies suggest ATL3 is a weak fusogen at best. Contrary to expectations, we demonstrate here that purified human ATL3 catalyzes efficient membrane fusion in vitro and is sufficient to sustain the ER network in triple knockout cells. Strikingly, ATL3 lacks any detectable C-terminal autoinhibition, like the invertebrate Drosophila ATL ortholog. Phylogenetic analysis of ATL C-termini indicates that C-terminal autoinhibition is a recent evolutionary innovation. We suggest that ATL3 is a constitutive ER fusion catalyst and that ATL1/2 autoinhibition likely evolved in vertebrates as a means of upregulating ER fusion activity on demand.