MOTIVATION: The increasing number of fully sequenced bacterial plasmids being annotated and catalogued has prompted the development of computational tools for comparing and classifying them. Existing approaches typically compare full-length DNA sequences (e.g. Mash, BLASTn, and ANI-based methods) or translated open reading frames (ORFs) (e.g. DIAMOND), with plasmid-level scores obtained by aggregating ORF-to-ORF similarities; however, they are either restricted to closely related plasmids or become computationally demanding in large-scale analyses. RESULTS: We describe pLAST (plasmid Language Analysis and Search Tool), a plasmid-search tool built using word2vec representations of protein-family content informed by local genomic context. Benchmarks indicate that pLAST outperforms nucleotide-based methods and performs comparably to DIAMOND in identifying functionally similar plasmids and compared with the widely used Mash, it achieves 26% and 24% improvements in detecting shared mating-pair formation system type and relaxase type, respectively. This performance scales to database searches across hundreds of thousands of sequences, as demonstrated using the precomputed PlasmidScope collection of ∼750 000 plasmids. Beyond global similarity, pLAST also returns per-ORF plasmid-plasmid alignments, enabling detection of shared functional modules. AVAILABILITY AND IMPLEMENTATION: pLAST is freely accessible as a web server at https://plast.lbs.cent.uw.edu.pl/ or https://plast.lbs.biol.uw.edu.pl/ and available as a Python module along with a precomputed database at https://github.com/labstructbioinf/pLAST for customized analysis.

MOTIVATION: The detection of homology through sequence comparison is a typical first step in the study of protein function and evolution. In this work, we explore the applicability of protein language models to this task. RESULTS: We introduce pLM-BLAST, a tool inspired by BLAST, that detects distant homology by comparing single-sequence representations (embeddings) derived from a protein language model, ProtT5. Our benchmarks reveal that pLM-BLAST maintains a level of accuracy on par with HHsearch for both highly similar sequences (with >50% identity) and markedly divergent sequences (with <30% identity), while being significantly faster. Additionally, pLM-BLAST stands out among other embedding-based tools due to its ability to compute local alignments. We show that these local alignments, produced by pLM-BLAST, often connect highly divergent proteins, thereby highlighting its potential to uncover previously undiscovered homologous relationships and improve protein annotation. AVAILABILITY AND IMPLEMENTATION: pLM-BLAST is accessible via the MPI Bioinformatics Toolkit as a web server for searching precomputed databases (https://toolkit.tuebingen.mpg.de/tools/plmblast). It is also available as a standalone tool for building custom databases and performing batch searches (https://github.com/labstructbioinf/pLM-BLAST).

The Rossmann fold enzymes are involved in essential biochemical pathways such as nucleotide and amino acid metabolism. Their functioning relies on interaction with cofactors, small nucleoside-based compounds specifically recognized by a conserved βαβ motif shared by all Rossmann fold proteins. While Rossmann methyltransferases recognize only a single cofactor type, the S-adenosylmethionine, the oxidoreductases, depending on the family, bind nicotinamide (nicotinamide adenine dinucleotide, nicotinamide adenine dinucleotide phosphate) or flavin-based (flavin adenine dinucleotide) cofactors. In this study, we showed that despite its short length, the βαβ motif unambiguously defines the specificity towards the cofactor. Following this observation, we trained two complementary deep learning models for the prediction of the cofactor specificity based on the sequence and structural features of the βαβ motif. A benchmark on two independent test sets, one containing βαβ motifs bearing no resemblance to those of the training set, and the other comprising 38 experimentally confirmed cases of rational design of the cofactor specificity, revealed the nearly perfect performance of the two methods. The Rossmann-toolbox protocols can be accessed via the webserver at https://lbs.cent.uw.edu.pl/rossmann-toolbox and are available as a Python package at https://github.com/labstructbioinf/rossmann-toolbox.

MOTIVATION: The wealth of protein structures collected in the Protein Data Bank enabled large-scale studies of their function and evolution. Such studies, however, require the generation of customized datasets combining the structural data with miscellaneous accessory resources providing functional, taxonomic and other annotations. Unfortunately, the functionality of currently available tools for the creation of such datasets is limited and their usage frequently requires laborious surveying of various data sources and resolving inconsistencies between their versions. RESULTS: To address this problem, we developed localpdb, a versatile Python library for the management of protein structures and their annotations. The library features a flexible plugin system enabling seamless unification of the structural data with diverse auxiliary resources, full version control and powerful functionality of creating highly customized datasets. The localpdb can be used in a wide range of bioinformatic tasks, in particular those involving large-scale protein structural analyses and machine learning. AVAILABILITY AND IMPLEMENTATION: localpdb is freely available at https://github.com/labstructbioinf/localpdb. Documentation along with the usage examples can be accessed at https://labstructbioinf.github.io/localpdb/.

MOTIVATION: Coiled coils are widespread protein domains involved in diverse processes ranging from providing structural rigidity to the transduction of conformational changes. They comprise two or more α-helices that are wound around each other to form a regular supercoiled bundle. Owing to this regularity, coiled-coil structures can be described with parametric equations, thus enabling the numerical representation of their properties, such as the degree and handedness of supercoiling, rotational state of the helices, and the offset between them. These descriptors are invaluable in understanding the function of coiled coils and designing new structures of this type. The existing tools for such calculations require manual preparation of input and are therefore not suitable for the high-throughput analyses. RESULTS: To address this problem, we developed SamCC-Turbo, a software for fully automated, per-residue measurement of coiled coils. By surveying Protein Data Bank with SamCC-Turbo, we generated a comprehensive atlas of ∼50 000 coiled-coil regions. This machine learning-ready dataset features precise measurements as well as decomposes coiled-coil structures into fragments characterized by various degrees of supercoiling. The potential applications of SamCC-Turbo are exemplified by analyses in which we reveal general structural features of coiled coils involved in functions requiring conformational plasticity. Finally, we discuss further directions in the prediction and modeling of coiled coils. AVAILABILITY AND IMPLEMENTATION: SamCC-Turbo is available as a web server (https://lbs.cent.uw.edu.pl/samcc_turbo) and as a Python library (https://github.com/labstructbioinf/samcc_turbo), whereas the results of the Protein Data Bank scan can be browsed and downloaded at https://lbs.cent.uw.edu.pl/ccdb. SUPPLEMENTARY INFORMATION: Supplementary data are available at Bioinformatics online.

While the slipknot topology in proteins has been known for over a decade, its evolutionary origin is still a mystery. We have identified a previously overlooked slipknot motif in a family of two-domain membrane transporters. Moreover, we found that these proteins are homologous to several families of unknotted membrane proteins. This allows us to directly investigate the evolution of the slipknot motif. Based on our comprehensive analysis of 17 distantly related protein families, we have found that slipknotted and unknotted proteins share a common structural motif. Furthermore, this motif is conserved on the sequential level as well. Our results suggest that, regardless of topology, the proteins we studied evolved from a common unknotted ancestor single domain protein. Our phylogenetic analysis suggests the presence of at least seven parallel evolutionary scenarios that led to the current diversity of proteins in question. The tools we have developed in the process can now be used to investigate the evolution of other repeated-domain proteins.

BACKGROUND: Protein repeats can confound sequence analyses because the repetitiveness of their amino acid sequences lead to difficulties in identifying whether similar repeats are due to convergent or divergent evolution. We noted that the patterns derived from traditional "dot plot" protein sequence self-similarity analysis tended to be conserved in sets of related repeat proteins and this conservation could be quantitated using a Jaccard metric. RESULTS: Comparison of these dot plots obviated the issues due to sequence similarity for analysis of repeat proteins. A high Jaccard similarity score was suggestive of a conserved relationship between closely related repeat proteins. The dot plot patterns decayed quickly in the absence of selective pressure with an expected loss of 50% of Jaccard similarity due to a loss of 8.2% sequence identity. To perform method testing, we assembled a standard set of 79 repeat proteins representing all the subgroups in RepeatsDB. Comparison of known repeat and non-repeat proteins from the PDB suggested that the information content in dot plots could be used to identify repeat proteins from pure sequence with no requirement for structural information. Analysis of the UniRef90 database suggested that 16.9% of all known proteins could be classified as repeat proteins. These 13.3 million putative repeat protein chains were clustered and a significant amount (82.9%) of clusters containing between 5 and 200 members were of a single functional type. CONCLUSIONS: Dot plot analysis of repeat proteins attempts to obviate issues that arise due to the sequence degeneracy of repeat proteins. These results show that this kind of analysis can efficiently be applied to analyze repeat proteins on a large scale.

MOTIVATION: Coiled coils are protein structural domains that mediate a plethora of biological interactions, and thus their reliable annotation is crucial for studies of protein structure and function. RESULTS: Here, we report DeepCoil, a new neural network-based tool for the detection of coiled-coil domains in protein sequences. In our benchmarks, DeepCoil significantly outperformed current state-of-the-art tools, such as PCOILS and Marcoil, both in the prediction of canonical and non-canonical coiled coils. Furthermore, in a scan of the human genome with DeepCoil, we detected many coiled-coil domains that remained undetected by other methods. This higher sensitivity of DeepCoil should make it a method of choice for accurate genome-wide detection of coiled-coil domains. AVAILABILITY AND IMPLEMENTATION: DeepCoil is written in Python and utilizes the Keras machine learning library. A web server is freely available at https://toolkit.tuebingen.mpg.de/#/tools/deepcoil and a standalone version can be downloaded at https://github.com/labstructbioinf/DeepCoil. SUPPLEMENTARY INFORMATION: Supplementary data are available at Bioinformatics online.

BACKGROUND: GmrSD is a modification-dependent restriction endonuclease that specifically targets and cleaves glucosylated hydroxymethylcytosine (glc-HMC) modified DNA. It is encoded either as two separate single-domain GmrS and GmrD proteins or as a single protein carrying both domains. Previous studies suggested that GmrS acts as endonuclease and NTPase whereas GmrD binds DNA. METHODS: In this work we applied homology detection, sequence conservation analysis, fold recognition and homology modeling methods to study sequence-structure-function relationships in the GmrSD restriction endonucleases family. We also analyzed the phylogeny and genomic context of the family members. RESULTS: Results of our comparative genomics study show that GmrS exhibits similarity to proteins from the ParB/Srx fold which can have both NTPase and nuclease activity. In contrast to the previous studies though, we attribute the nuclease activity also to GmrD as we found it to contain the HNH endonuclease motif. We revealed residues potentially important for structure and function in both domains. Moreover, we found that GmrSD systems exist predominantly as a fused, double-domain form rather than as a heterodimer and that their homologs are often encoded in regions enriched in defense and gene mobility-related elements. Finally, phylogenetic reconstructions of GmrS and GmrD domains revealed that they coevolved and only few GmrSD systems appear to be assembled from distantly related GmrS and GmrD components. CONCLUSIONS: Our study provides insight into sequence-structure-function relationships in the yet poorly characterized family of Type IV restriction enzymes. Comparative genomics allowed to propose possible role of GmrD domain in the function of the GmrSD enzyme and possible active sites of both GmrS and GmrD domains. Presented results can guide further experimental characterization of these enzymes.

BACKGROUND: SPOUT methyltransferases (MTases) are a large class of S-adenosyl-L-methionine-dependent enzymes that exhibit an unusual alpha/beta fold with a very deep topological knot. In 2001, when no crystal structures were available for any of these proteins, Anantharaman, Koonin, and Aravind identified homology between SpoU and TrmD MTases and defined the SPOUT superfamily. Since then, multiple crystal structures of knotted MTases have been solved and numerous new homologous sequences appeared in the databases. However, no comprehensive comparative analysis of these proteins has been carried out to classify them based on structural and evolutionary criteria and to guide functional predictions. RESULTS: We carried out extensive searches of databases of protein structures and sequences to collect all members of previously identified SPOUT MTases, and to identify previously unknown homologs. Based on sequence clustering, characterization of domain architecture, structure predictions and sequence/structure comparisons, we re-defined families within the SPOUT superfamily and predicted putative active sites and biochemical functions for the so far uncharacterized members. We have also delineated the common core of SPOUT MTases and inferred a multiple sequence alignment for the conserved knot region, from which we calculated the phylogenetic tree of the superfamily. We have also studied phylogenetic distribution of different families, and used this information to infer the evolutionary history of the SPOUT superfamily. CONCLUSION: We present the first phylogenetic tree of the SPOUT superfamily since it was defined, together with a new scheme for its classification, and discussion about conservation of sequence and structure in different families, and their functional implications. We identified four protein families as new members of the SPOUT superfamily. Three of these families are functionally uncharacterized (COG1772, COG1901, and COG4080), and one (COG1756 represented by Nep1p) has been already implicated in RNA metabolism, but its biochemical function has been unknown. Based on the inference of orthologous and paralogous relationships between all SPOUT families we propose that the Last Universal Common Ancestor (LUCA) of all extant organisms contained at least three SPOUT members, ancestors of contemporary RNA MTases that carry out m1G, m3U, and 2'O-ribose methylation, respectively. In this work we also speculate on the origin of the knot and propose possible 'unknotted' ancestors. The results of our analysis provide a comprehensive 'roadmap' for experimental characterization of SPOUT MTases and interpretation of functional studies in the light of sequence-structure relationships.