Communications Chemistry, Published online: 18 September 2026; doi:10.1038/s42004-026-02198-xSeveral multiple sequence alignment (MSA)-perturbing strategies have been shown to enhance AlphaFold2 (AF2) sampling of alternative protein states. In this work, the authors evaluate these strategies on AlphaFold3 (AF3), showing that the MSA-perturbing strategies enhance AF3 sampling of alternative states, and that different choices of amino acid masks can improve column-masked AF3 sampling for specific targets.

ByteDance's newly independent AI-biology unit reportedly raised $290M at a $1.5B valuation -- but its most concrete product is Protenix, a free model its own paper says beats DeepMind's AlphaFold3.

An AI system trained on more than 20,000 protein structures from pharmaceutical companies outperforms AlphaFold-like models that use only public data.

Legumes convert atmospheric nitrogen into ammonium through symbiotic bacteria housed in root nodules, yet the molecular interactions between rhizobial and host proteins inside nodules remain poorly understood. Here we employed AlphaFold3 to construct a cross-kingdom interactome between Medicago truncatula and its symbiont Sinorhizobium meliloti. Screening more than 217,000 protein pairs yielded 7,137 putative interactions, providing a valuable resource for the broader symbiosis community. Within this network, we focused on DEFECTIVE IN NITROGEN FIXATION 2 (DNF2), a host protein required for rhizobial persistence within nodules. We showed that DNF2 localizes to the peribacteroid space and associates with previously uncharacterized secreted rhizobial proteins (SRPs), suggesting it may function as a hub for host-symbiont communication. Notably, knockout of two DNF2-interacting proteins, SRP86 and SRP485, results in white, nitrogen-fixation-deficient nodules with abnormal symbiosomes and elevated expression of senescence-associated genes, closely phenocopying the dnf2 loss-of-function mutant. Together, our findings define a DNF2-SRP molecular framework underlying symbiotic accommodation, and illustrate the potential of AI-guided interactome mapping to uncover molecular mechanisms of plant-microbe interactions with relevance to sustainable agriculture.

Identifying off-target interactions of approved drugs is important to anticipate side effects and uncover repurposing opportunities. Computational pipelines combining structural homology, structure pr...

We introduce ABLE, a benchmark for evaluating LLM agents' ability to use biological AI models (BAIMs), such as ProteinMPNN and AlphaFold3, in dual-use protein design workflows. ABLE assesses agent performance through a set of tasks spanning structure retrieval, sequence generation, and design valid…

DeepMind winds down its Nobel-winning protein team as researchers move to Gemini and to Anthropic, with AlphaFold3 staying open

The SBGrid Consortium is an innovative global research computing group operated out of Harvard Medical School. SBGrid provides the global structural biology community with support for research…

Abstract page for arXiv paper 2609.05818: Agentic BAIM-LLM Evaluation (ABLE): Benchmarking LLM Use of Protein Design Tools

AlphaFold 3 inference pipeline. Contribute to sokrypton/alphafold3 development by creating an account on GitHub.

Legumes convert atmospheric nitrogen into ammonium through symbiotic bacteria housed in root nodules, yet the molecular interactions between rhizobial and host proteins inside nodules remain poorly un...