Andrew D White

Jon M Laurent, Albert Bou, Michael Pieler et al.

Optimism for accelerating scientific discovery with AI continues to grow. Current applications of AI in scientific research range from training dedicated foundation models on scientific data to agentic autonomous hypothesis generation systems to AI-driven autonomous labs. The need to measure progress of AI systems in scientific domains correspondingly must not only accelerate, but increasingly shift focus to more real-world capabilities. Beyond rote knowledge and even just reasoning to actually measuring the ability to perform meaningful work. Prior work introduced the Language Agent Biology Benchmark LAB-Bench as an initial attempt at measuring these abilities. Here we introduce an evolution of that benchmark, LABBench2, for measuring real-world capabilities of AI systems performing useful scientific tasks. LABBench2 comprises nearly 1,900 tasks and is, for the most part, a continuation of LAB-Bench, measuring similar capabilities but in more realistic contexts. We evaluate performance of current frontier models, and show that while abilities measured by LAB-Bench and LABBench2 have improved substantially, LABBench2 provides a meaningful jump in difficulty (model-specific accuracy differences range from -26% to -46% across subtasks) and underscores continued room for performance improvement. LABBench2 continues the legacy of LAB-Bench as a de facto benchmark for AI scientific research capabilities and we hope that it continues to help advance development of AI tools for these core research functions. To facilitate community use and development, we provide the task dataset at https://huggingface.co/datasets/futurehouse/labbench2 and a public eval harness at https://github.com/EdisonScientific/labbench2.

Rainier Barrett, Shaoyi Jiang, Andrew D White

Bayesian network models are finding success in characterizing enzyme-catalyzed reactions, slow conformational changes, predicting enzyme inhibition, and genomics. In this work, we apply them to statistical modeling of peptides by simultaneously identifying amino acid sequence motifs and using a motif-based model to clarify the role motifs may play in antimicrobial activity. We construct models of increasing sophistication, demonstrating how chemical knowledge of a peptide system may be embedded without requiring new derivation of model fitting equations after changing model structure. These models are used to construct classifiers with good performance (94% accuracy, Matthews correlation coefficient of 0.87) at predicting antimicrobial activity in peptides, while at the same time being built of interpretable parameters. We demonstrate use of these models to identify peptides that are potentially both antimicrobial and antifouling, and show that the background distribution of amino acids could play a greater role in activity than sequence motifs do. This provides an advancement in the type of peptide activity modeling that can be done and the ease in which models can be constructed.