TRILL: ORCHESTRATING MODULAR DEEP-LEARNING WORKFLOWS FOR DEMOCRATIZED, SCALABLE PROTEIN ANALYSIS AND ENGINEERING.

Deep-learning models have been rapidly adopted by many fields, partly due to the deluge of data humanity has amassed. In particular, the petabases of biological sequencing data enable the unsupervised training of protein language models that learn the "language of life." However, due to their prohibitive size and complexity, contemporary deep-learning models are often unwieldy, especially for scientists with limited machine learning backgrounds. TRILL (TRaining and Inference using the Language of Life) is a platform for creative protein design and discovery. Leveraging several state-of-the-art models such as ESM-2, DiffDock, and RFDiffusion, TRILL allows researchers to generate novel proteins, predict 3-D structures, extract high-dimensional representations of proteins, functionally classify proteins and more. What sets TRILL apart is its ability to enable complex pipelines by chaining together models and effectively merging the capabilities of different models to achieve a sum greater than its individual parts. Whether using Google Colab with one GPU or a supercomputer with hundreds, TRILL allows scientists to effectively utilize models with millions to billions of parameters by using optimized training strategies such as ZeRO-Offload and distributed data parallel. Therefore, TRILL not only bridges the gap between complex deep-learning models and their practical application in the field of biology, but also simplifies the orchestration of these models into comprehensive workflows, democratizing access to powerful methods. Documentation: https://trill.readthedocs.io/en/latest/home.html.

Organizing the Global Diversity of Microviruses.

Microviruses encompass an astonishing array of small, single-stranded DNA phages that, due to the surge in metagenomic surveys, are now known to be prevalent in most environments. Current taxonomy concedes the considerable diversity within this lineage to a single family (the Microviridae), which has rendered it difficult to adequately and accurately assess the amount of variation that actually exists within this group. We amassed and curated the largest collection of microviral genomes to date and, through a combination of protein-sharing networks and phylogenetic analysis, discovered at least three meaningful taxonomic levels between the current ranks of family and genus. When considering more than 13,000 microviral genomes from recognized lineages and as-yet-unclassified microviruses in metagenomic samples, microviral diversity is better understood by elevating microviruses to the level of an order that consists of three suborders and at least 19 putative families, each with their respective subfamilies. These revisions enable fine-scale assessment of microviral dynamics: for example, in the human gut, there are considerable differences in the abundances of microviral families both between urban and rural populations and in individuals over time. In addition, our analysis of genome contents and gene exchange shows that microviral families carry no recognizable accessory metabolic genes and rarely, if ever, engage in horizontal gene transfer across microviral families or with their bacterial hosts. These insights bring microviral taxonomy in line with current developments in the taxonomy of other phages and increase the understanding of microvirus biology. IMPORTANCE Microviruses are the most abundant single-stranded DNA phages on the planet and an important component of the human gut virome. And yet, productive research into their biology is hampered by the inadequacies of current taxonomic ordering: microviruses are lumped into a single family and treated as a monolithic group, thereby obscuring the extent of their diversity and resulting in little comparative research. Our investigations into the diversity of microviruses define numerous groups, most lacking any isolated representatives, and point toward high-value targets for future research. To expedite microvirus discovery and comparison, we developed a pipeline that enables the fast and facile sorting of novel microvirus genomes into well-defined taxonomic groups. These improvements provide new insights into the biology of microviruses and emphasize fundamental differences between these miniature phages and their large, double-stranded DNA phage competitors.

Defensive hypervariable regions confer superinfection exclusion in microviruses.

Single-stranded DNA phages of the family Microviridae have fundamentally different evolutionary origins and dynamics than the more frequently studied double-stranded DNA phages. Despite their small size (around 5 kb), which imposes extreme constraints on genomic innovation, they have adapted to become prominent members of viromes in numerous ecosystems and hold a dominant position among viruses in the human gut. We show that multiple, divergent lineages in the family Microviridae have independently become capable of lysogenizing hosts and have convergently developed hypervariable regions in their DNA pilot protein, which is responsible for injecting the phage genome into the host. By creating microviruses with combinations of genomic segments from different phages and infecting Escherichia coli as a model system, we demonstrate that this hypervariable region confers the ability of temperate Microviridae to prevent DNA injection and infection by other microviruses. The DNA pilot protein is present in most microviruses, but has been recruited repeatedly into this additional role as microviruses altered their lifestyle by evolving the ability to integrate in bacterial genomes, which linked their survival to that of their hosts. Our results emphasize that competition between viruses is a considerable and often overlooked source of selective pressure, and by producing similar evolutionary outcomes in distinct lineages, it underlies the prevalence of hypervariable regions in the genomes of microviruses and perhaps beyond.