Google DeepMind’s Cheminformatics Skills bring RDKit-based compound curation, an interactive molecular viewer, matched molecular pair analysis and predictive modelling into Antigravity. A lung-cancer drug case study demonstrates literature-to-design workflows, including proposed macrocycles and synthesis routes. Install the Science plugin through Antigravity’s in-app marketplace; Google’s Science page also links its public skills repository. The proposed molecules remain computational hypotheses requiring synthesis and experimental validation.
Updates
PhageFlare: deep-learning tool to annotate phage proteins that structure-similarity searches miss (bioRxiv preprint)
A Hong Kong Baptist University team (corresponding author Lu Zhang) posted the bioRxiv preprint "Illuminating phage dark matter beyond the reach of structural alignments with PhageFlare" on 29 October 2026, not yet peer reviewed. PhageFlare is a deep-learning framework (frozen ESM-C 600M protein language model, 45 functional labels, optional genomic context) for predicting the functions of phage proteins that have no significant structural match to annotated references. The authors report macro-averaged F1 of 0.871, 0.889 and 0.738 on three structure-disjoint benchmarks, beating the strongest baseline, and say it flagged lysogenic potential in 1,866 prophage-designated genomes that BACPHLIP had classed as lytic. Code is on GitHub (MIT licence) and the model checkpoint is on Hugging Face.
Arc/Stanford: Evo 2 can design bacteriophage genomes from a handful of examples in the prompt
Gary Brixi and Brian Hie's teams report that the genome language model Evo 2 can learn from a few examples placed in its prompt (in-context learning, no retraining) and use them to generate proteins, de novo antibodies, CRISPR-Cas9 loci, mitochondrial genomes and entire bacteriophage genomes. Feeding the model's own synthetic sequences back into the prompt (a 'chain-of-thought') helps reach harder design goals. The work is the bioRxiv preprint "Efficient biological design via genomic in-context learning", posted 29 October 2026 and not yet peer reviewed.
DeepCellMorpho benchmarks self-supervised models for Cell Painting
DeepCellMorpho compares four self-supervised models trained on roughly 1.77 million Cell Painting images. In the Nature Methods study, DINO and DINOv2 performed strongly on compound and mechanism-of-action retrieval, using image representations without cell segmentation or hand-engineered features. The authors provide inference code, pretrained models and embeddings; model weights and embeddings are restricted to noncommercial use. Cross-laboratory batch effects and interpretability remain limitations.
Global CRISPR SpacerDB links metagenomic spacers with viruses and plasmids
JGI’s Global CRISPR Spacer Database provides downloadable spacer sequences, virus/plasmid matches and example notebooks for exploring microbial interactions. Its accompanying Nature Microbiology study collected 794 million spacers from 367,609 metagenomes. The study cautions that CRISPR targeting does not always establish a productive virus–host relationship. The database predates the October 2026 paper.
Morgen:Rapport: Lars Fieseler on phages and food safety
In this 42-minute German-language podcast, ZHAW microbiologist Lars Fieseler speaks with Jean-Daniel Strub about phage specificity, applications in food production and how phage resistance differs from antibiotic resistance. The episode was released on October 27, 2026.
Thirteen quick tips for bacteriophage genome annotation
An open-access practical guide to annotating tailed phage genomes, from assembly checks and phage-aware gene calling to conservative functional assignments. The authors explain why automated outputs need manual review, how gene context and protein structure can strengthen predictions, and why an honest hypothetical-protein label is preferable to an unsupported function that spreads through databases. A tool-and-database table, an annotator checklist and a companion HTML workflow make the guide useful alongside an annotation project or workshop.
Ottawa case report follows phage-treated joint infection for two years
A two-year follow-up preprint from Azad Laboratory reports sustained clinical control of a multidrug-resistant Staphylococcus epidermidis hip prosthetic-joint infection after personalized phage therapy alongside antibiotics and surgery. Samples collected during later surgery for a fall-related hip dislocation were negative by bacterial culture and broad-range 16S rRNA sequencing. The patient remains on suppressive daptomycin, so this single case cannot establish an antibiotic-free cure or isolate the contribution of phages. The authors also propose standardized endpoints for future phage-therapy studies in joint infections.
Alex Rives on virtual cells, measurement and Biohub’s $1.8B data initiative
Biohub Head of Science Alex Rives discusses the limits of virtual-cell models, the need for better measurement and data, Biohub’s $1.8B data initiative, cross-scale biology and experimental validation. Rives says an information-level model of the cell may be attainable within about five years.
NLM institutional training grants in biomedical informatics and data science
NLM’s RFA-LM-28-001 supports U.S. institutions providing predoctoral and postdoctoral research training in biomedical informatics and data science, including AI and machine learning. This is an institutional grant, not an individual fellowship. Applications can be submitted from January 11, 2027 and are due February 11 at 5 p.m. applicant-local time. Trainees must be U.S. citizens, nationals or permanent residents.
Preprint maps host determinants of mycobacteriophage infection with genome-wide CRISPRi
A Phage Foundry preprint screens a genome-wide CRISPRi library of Mycobacterium smegmatis against 70 mycobacteriophages. Outer-membrane Msp porins emerged as the dominant candidate receptor family for 46 phages, and determinant usage did not follow phage phylogeny.
Phage prohead proteases activate CBASS antiphage immunity (Science)
A Science paper by Samuel J. Hobbs and Philip J. Kranzusch reports that phage prohead protease activity activates the CD-NTase enzymes of bacterial CBASS defence: the authors reconstitute CBASS recognition of phage T4 in vitro, identify cleavage of a CD-NTase activation loop as the trigger of immune signal synthesis, and show that phage prohead proteases activate CBASS in vivo, explaining how immunity is initiated late during infection. Structure-based analyses identify activation loops in clade A, D and G CD-NTases.
TattleTail identifies canonical pyocin gene clusters in Pseudomonas aeruginosa
TattleTail is a rule-based bioinformatic pipeline for identifying canonical pyocin-encoding gene clusters in Pseudomonas aeruginosa. It targets the characteristic trpE–trpG-associated loci, combining conserved pyocin genes with exclusion of canonical phage features to distinguish these clusters from prophage regions. It complements existing prophage prediction tools.
Phylogeny-agnostic machine learning predicts strain-level phage–host interactions from genomes (Nature Microbiology)
A Nature Microbiology study from Lawrence Berkeley National Laboratory, UC Berkeley and Penn State presents a phylogeny-agnostic machine-learning framework that predicts strain-level phage–host interactions across diverse bacterial genera from genome sequences alone, matching species-specific methods (AUROC 0.67–0.94) across six datasets (115,037 interactions, 949 bacterial strains, 518 phages). Experimental validation of 1,240 predicted E. coli phage–host interactions gave an AUROC of 0.84, and model-guided cocktail design reached up to 97.5% bacterial coverage with five phages.
HIDEN-SEQ: transposon-insertion sequencing that links phage genes to selectable phenotypes (Nature Microbiology)
A Nature Microbiology study from ETH Zürich, University Hospital Basel, the University of Basel and the University of Lausanne presents HIDEN-SEQ (hidden Acr-enabled transposon-insertion sequencing), which systematically links viral genes to selectable phenotypes. Using phage T4 it reproduced the established gene essentiality map, and it is portable to non-model phages, revealing conditionally essential genes in several bacterial hosts and conditions, including previously unknown antidefence factors matched to specific antiviral defences.
Defining the essential genome of diverse phages with phage Tn-seq
Phage Tn-seq is a transposon-based approach to studying phage gene essentiality and engineering.
Call for papers: Virologica Sinica special issue "Phage Therapy and Beyond", submissions due 30 April 2027
Virologica Sinica has opened a special issue, "Phage Therapy and Beyond", marking 110 years since d'Herelle named and began using phage therapy. It welcomes research on clinical phage therapy, resistance, structure, genomics, endolysins and related topics; the submission deadline is 30 April 2027.
HydroPhage: a hyaluronan hydrogel that delivers phage over one week in a mouse wound infection model (Nature Communications)
A Nature Communications study from Stanford University and collaborators used a mouse model of established Pseudomonas aeruginosa wound infection to compare phage route, dose and frequency: topical, but not systemic, delivery reduced wound bacterial burden, and phage resistance emerged rapidly, suggesting a benefit from high, repeated doses. The team then developed HydroPhage, a hyaluronan-based hydrogel with dynamic covalent crosslinks that delivers high-titer phage over one week.
Prophage-encoding engineered bacteria enable prophylactic lytic phage therapy for enteric infection in mice
Engineered non-pathogenic E. coli producing lytic phages protected mice against Salmonella infection in a preclinical proof-of-concept study.
BFVD v3 expands viral protein structure predictions to 5.8 million
BFVD v3 expands the Big Fantastic Virus Database to 5,776,417 predicted viral protein structures, a 16.4-fold increase over v2. Released September 15, 2026, it covers individual viral sequences from UniProt 2025_03 and provides downloadable structures, metadata and a Foldseek-searchable database under CC BY 4.0. It supports viral protein structure comparisons, including phage research; these are computational predictions, not experimentally determined structures.