Search bioRxiv⌕ Search

Biology subjects

Omasa, T.

Publications and source records attributed to Omasa, T..

2 recordsLinked to original sources

Harness engineering enables a multi-agent LLM system to generate cross-domain hypotheses beyond single-investigator reach

Large language models (LLMs) are increasingly used to generate scientific hypotheses, and a more capable model is widely assumed to drive output quality. We tested this with a biology-focused multi-agent system coupling an LLM to five Model Context Protocol servers exposing 66 external tools. Upgrading the Synthesizer and Critic from Claude Sonnet 4.6 to Opus 4.8 initially appeared to improve quality, count, and reliability. Under a fully matched harness (identical prompts, schema, hypothesis count, and queries), the differences vanished: scores were indistinguishable on an iPSC-reprogramming task and favoured Sonnet 4.6 on a CHO cell-line-engineering task (18.4 vs 17.4; Welch t = 2.89, p < 0.05). The dominant, controllable driver of quality is therefore the prompting harness, not the model tier. We then introduce the ABCD-Stack (contract-gated prompting, diversity-first generation, adversarial refinement, and cross-family evaluation), which builds these properties in as architecture and shifts outputs toward explicitly falsifiable, causally framed hypotheses beyond the reach of a single-domain investigator. The hypotheses are falsifiable by construction but not yet experimentally validated.

bioinformatics↗

Database Chemistry for Genomics-Based Safety and Quality Evaluation of Biologics

Genomics-based safety and quality evaluation studies are advancing the bioindustry by enhancing various aspects, including viral safety, host cell protein (HCP) control, product heterogeneity control, cellular heterogeneity control, and process reproducibility. High-throughput instruments and genome-scale databases are essential in genomics, with the reference genome sequence being the most critical database. The completeness and accuracy of these genome sequences depend on DNA quality, sequencing instruments, read coverage, and assembly strategies. Significant efforts are being made to perfect genome assembly and continuously improve it. However, the quantitative impact of reference genome sequence accuracy on the safety of biologics is not yet fully understood. In this study, we compared and benchmarked six Chinese hamster genomes, including four newly sequenced genomes derived from Chinese hamster cell lines, from an industrial perspective. We also developed database assembly techniques to enhance the safety of biologics. We recommend using two or more independent reference genomes for viral safety studies. For HCP control, we suggest using protein sequences in which trypsin degradation peptides that overlap with high-risk proteins should be masked and unified. Additionally, we can predict microenvironments using single-cell transcriptome data. In bioengineering processes, any nucleotide samples have potential commercial benefits.

bioengineering↗