Research agents that show their work.

XS1 Biosciences researches scientific agents that read the literature, write and run analysis code, and rank hypotheses and draft experiment plans — with every step recorded and a researcher deciding what happens next.

Illustrative runResearch question

Which candidate genes might regulate a stress-response pathway? Use the published literature and one public expression dataset.

Breaks the question into a literature search, an expression analysis and a ranking.

Plan

  1. 1Search the literature for known regulators of the pathway
  2. 2Test differential expression, controlling for batch
  3. 3Rank candidate genes by the combined evidence

01What agents do

The dry-lab work around an experiment.

Agents take on the reading, coding and bookkeeping that surround an experiment, so researchers spend their time on the decisions.

  • Literature review

    Reading and summarizing published work and databases, with each claim linked to the source it came from.

    • Search
    • Summaries
    • Citations
  • Dry-lab analysis

    Writing and running analysis code in a sandbox: sequencing (NGS) analysis such as alignment, variant calling and expression quantification, and the statistics that follow.

    • NGS analysis
    • Statistics
    • Reproducible code
  • Hypotheses

    Proposing candidate explanations and ranking them by the evidence for and against each, including the boring explanations a control can rule out.

    • Generation
    • Ranking
    • Controls
  • Experiment planning

    Drafting protocols, controls and troubleshooting steps for an experiment, for a researcher to review, change or reject.

    • Protocol drafts
    • Controls
    • Troubleshooting

02How they work

Written like a protocol.

The same five rules apply to every research agent XS1 Biosciences builds.

  1. Rule 01Sources for every claimStatements link to the papers, databases or results they rest on, so a reader can check them.
  2. Rule 02Reproducible analysisCode, inputs and versions are kept with each result, so an analysis can be rerun and audited.
  3. Rule 03Scoped toolsAgents use the tools, databases and compute a project allows, and nothing else.
  4. Rule 04Uncertainty statedRankings come with the evidence for and against each option, and gaps are named rather than filled.
  5. Rule 05People decideResearchers decide what is tested in the lab, what is reported and what happens next.

Agents plan and analyze; they do not decide what reaches the bench. They are built to stay within a project's agreed scope and to decline requests that raise biosecurity concerns.

Work with XS1 Biosciences

Bring XS1 Biosciences a research workflow.

A literature question, an analysis pipeline, a recurring dry-lab task: describe it in general terms and we will follow up.

Please do not send unpublished data, sequences or other confidential material through this website.