Rita Pancsa

Background: Positive selection shapes protein function by favoring amino acid substitutions that increase fitness. Although numerous studies have identified positively selected genes (PSGs), the structural principles underlying residue-level positive selection remain incompletely understood and previous studies have produced conflicting conclusions, particularly regarding the preferential location of positively selected residues (PSRs) within intrinsically disordered regions. In order to resolve these issues and to better understand the driving forces underlying positive selection in human proteins, we performed a comprehensive analysis on a manually curated, high quality collection of PSGs and PSRs as well as on a large-scale PSR dataset restricted to proteins with experimental structures. Results: Human PSRs were significantly enriched in secreted and cell membrane proteins and preferentially localized to extracellular regions, solvent-exposed protein surfaces, coil structures, and protein-protein interaction interfaces, whereas neither domains nor intrinsically disordered regions showed an enrichment. Spatial clustering analysis reinforced that PSRs accumulate within localized surface patches. Integration of host-pathogen interaction data highlighted that adaptive changes preferentially affect extracellular molecular recognition surfaces, recurrently appearing in certain pathogen-interacting membrane protein families and particular domain types. Additionally, PSRs showed an enrichment in the residues of human-pathogen interaction interfaces that are in direct contact with pathogenic proteins, implying that neutralizing pathogen attacks is one of the driving forces behind the adaptive evolution of human proteins. By analysing human variation data we found that PSRs are enriched in benign but depleted in pathogenic substitutions, extending previous observations that positively selected genes contain elevated levels of missense variation. Notably, PSGs of the large-scale dataset were also enriched among clinical-stage drug targets, suggesting a potential link between positive selection and pharmacological relevance. Conclusion: Together, our results provide a refined structural model of adaptive evolution in human proteins and identify extracellular exposed molecular recognition surfaces and human-pathogen interaction interfaces as recurrent hotspots of positive selection. Our results point to the direction that precise identification of human PSRs could delineate important host-pathogen interfaces that exert evolutionary selection pressure on humans and appoint novel targets with therapeutic potential.

The androgen receptor (AR) is a transcription factor whose overactivation is a primary driver of prostate cancer. Although AR interactions with several long non-coding RNAs (lncRNAs) have been implicated in castration-resistant prostate cancer, their underlying molecular mechanisms and functional consequences remain poorly understood. Here, we identify the N-terminal 37 residues of the intrinsically disordered AR N-terminal domain as the primary RNA-binding region that mediates selective interactions with the lncRNAs HOTAIR and SLNCR1. Residue-level mapping and mutational analysis define Y11, R13, and Q24 as key determinants of RNA recognition. This RNA-binding region partially overlaps with the F23QNLF27 motif, previously shown to mediate N/C interdomain communication with the ligand-binding domain through folding upon binding. We observe a partner-dependent binding mode in which this motif remains dynamically disordered upon RNA binding. LncRNA binding promotes phase separation of the N-terminal domain, indicating that lncRNAs upregulated in late-stage prostate cancer may lower the threshold for AR condensate formation and contribute to ligand-independent AR signaling. LncRNAs modulate communication between the N-terminal and ligand-binding domains within condensates, suggesting that lncRNA binding may tune hormone-dependent full-length AR signaling. These findings provide a mechanistic framework for AR-lncRNA regulation, laying the groundwork for future therapeutic strategies against advanced prostate cancer.

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.

Short linear motifs (SLiMs) are small interaction modules within intrinsically disordered regions of proteins that interact with specific domains, and thereby regulate diverse biological processes. The limited information in SLiM definitions leads to frequent false positive hits in computational and experimental SLiM identification methods. Here, we refined the annotations of known motif classes to create a high-quality training dataset, and developed SLiMMine, a deep learning-based method to identify SLiMs in the human proteome. Relying on protein embeddings, SLiMMine outperforms similar approaches in identifying functional SLiMs, while eliminating ~80% of the pattern matching-based hits as false-positives. Furthermore, it can uncover novel SLiMs that do not fit known motif classes purely based on sequence features. Finally, narrowing the broad interactor-domain definitions of SLiM classes to specific partners enables more precise linking of predicted SLiMs to known protein-protein interactions. SLiMMine is available as a user-friendly, multi-purpose web server at https://slimmine.pbrg.hu/.

SH3 domains are widespread protein modules that mostly bind to proline-rich short linear motifs (SLiMs). Most known SH3 domain-motif interactions and canonical or non-canonical recognition specificities are described for individual SH3 domains. Although cooperation and coordinated motif binding between tandem SH3 domains has already been described for members of the p47^phox-related protein family, individual cases have never been collected and analyzed collectively, which precluded the definition of the binding preferences and targeted discovery of further instances. Here, we apply an integrative approach that includes data collection, curation, bioinformatics analyses and state-of-the-art structure prediction methods to fill these gaps. A search of the human proteome with the sequence signatures of SH3 tandemization and follow-up structure analyses suggest that SH3 tandemization could be specific for this family. We define the optimal binding preference of tandemly arranged SH3 domains as [PAVIL]PPR[PR][^DE][^DE] and propose potential new instances of this SLiM among the family members and their binding partners. Structure predictions suggest the possibility of a novel, reverse binding mode for certain motif instances. In all, our comprehensive analysis of this unique SH3 binding mode enabled the identification of novel, interesting tandem SH3-binding motif candidates with potential therapeutic relevance.

DisProt (https://disprot.org/) is an open database integrating experimental evidence on intrinsically disordered proteins (IDPs), intrinsically disordered regions (IDRs), and their functions. Over the past two years, the database has grown over 20%, now comprising 3201 IDPs and 13 347 pieces of evidence, including over 1500 new structural state annotations and >1300 new function annotations. DisProt has systematically adopted the Minimum Information About Disorder Experiments (MIADE) guidelines, more than doubling annotations with experimental details and improving the interpretability of disorder-related experiments. The website has evolved into a hybrid knowledgebase and deposition system, introducing a Deposition Page that allows direct submissions by external users. Through BLAST-based homology propagation in MobiDB, DisProt disorder regions and linear interacting peptides have been extended from hundreds to hundreds of thousands of proteins across >11 000 organisms. This new release marks a paradigm shift by integrating computational predictions as valid evidence and introducing major updates and restructuring of the IDP Ontology, enhancing accuracy, interoperability, and semantic clarity. DisProt continues to support community engagement through training resources together with DisTriage, an AI-based literature triage tool, providing curators with regularly updated lists of prioritized publications.

Leronlimab, a humanized monoclonal antibody directed to the CCR5 receptor, demonstrated anti-tumor activity in humanized NSG mouse models of both heterotopic and orthotopic SW480 human colon carcinoma. The anti-tumor effect was lost in non-humanized mice, suggesting a critical role for engrafted human leukocytes. Development of xGVHD was delayed in both tumor-bearing and non-tumor-bearing mice that had been treated with leronlimab. Leronlimab induced human CD4+CD25+ cells in the peripheral blood of tumor-bearing (1.5-fold increase, p=0.016) and non-tumor-bearing animals (2.2-fold, p=0.0038). Leronlimab reduced CD4+CD25- cells (1.84-fold reduction, p=0.033). In the orthotopic colon cancer model, lung metastatic burden was decreased ~87% in leronlimab-treated mice compared to IgG-treated animals (p=0.012). Blockade of CCR5 signaling reduced peri-tumoral vessel formation feeding the tumor, 62% (p=0.013). As CD4+CD25+ T cells, play a crucial role in immune checkpoint inhibitor (ICI) responses we assessed CCR5/immune checkpoint gene expression in human colon cancer. CCR5 and its ligands were overexpressed in human colon cancer, correlating with increased immune checkpoint gene expression in both microsatellite stable (MSS) and instable (MSI) patients.

Numerous cellular processes rely on biomolecular condensates formed through liquid–liquid phase separation (LLPS). Recently, it has become evident that somatic mutations can interfere with or over-activate the formation of phase-separated condensates. Here, we set out to systematically study the connection between cancer and biological condensation, specifically mapping the extent to which LLPS is affected in cancer and understanding the molecular pathomechanisms and therapeutic consequences of mutations affecting LLPS scaffolds. We identify both known and novel combinations of molecular functions that are specific to oncogenic fusion proteins and thus have a high potential for driving tumorigenesis. Protein regions driving condensate formation show an increased association with DNA- or chromatin-binding domains of transcription regulators within oncogenic fusion proteins, indicating a common molecular mechanism underlying several soft tissue sarcomas and hematologic malignancies where phase-separation-prone oncogenic fusion proteins form abnormal condensates along the DNA and thereby dysregulate gene expression programs. We find that proteins initiating LLPS are frequently implicated in somatic cancers, even surpassing their involvement in neurodegeneration. Our data shows that cancer-driving LLPS scaffolds tend to be potent oncogenes, giving rise to dominant phenotypes and lacking targeting options by current FDA-approved drugs. Finding the currently missing drugs to shut down oncogenic fusion proteins, to disrupt the condensation enabled by them, and to offset their downstream effects could provide cancer drugs widely applicable to diverse cancer incidences previously defying standard treatments.

SH3 domains are widespread protein modules that mostly bind to proline-rich short linear motifs (SLiMs). Most known SH3 domain-motif interactions and canonical or non-canonical recognition specificities are described for individual SH3 domains. Although cooperation and coordinated motif binding between tandem SH3 domains has already been described for members of the p47 ^phox -related protein family, individual cases have never been collected and analyzed collectively, which precluded the definition of the binding preferences and targeted discovery of further instances. Here, we apply an integrative approach that includes data collection, curation, bioinformatics analyses and state-of-the-art structure prediction methods to fill these gaps. We define the optimal binding preference of tandemly arranged SH3 domains as [PAVIL]PPR[PR][^DE][^DE], and propose potential new instances of this SLiM among the family members and their binding partners. Structure predictions suggest the possibility of a novel, reverse binding mode for certain motif instances. A search of the human proteome with the sequence signatures of SH3 tandemization and follow-up structure analyses suggest that SH3 tandemization could be specific for this family. In all, our comprehensive analysis of this unique SH3 binding mode enabled description of the binding preference and identification of novel interesting cases proposed for experimental validation.