Load your data
Drop in count matrices, proteomics tables, VCF files, or metabolomics peak lists. The system checks that files are valid and samples match up. Connect your own cloud storage or use ours.
Data characterization
The agent profiles your dataset: sample distributions, missing values, potential batch effects, outliers. You see a PCA plot and summary statistics. This is the baseline before any analysis.
Literature research
The agent searches published papers for methods that worked on similar data. It pulls citations, checks sample sizes, notes limitations. You get a summary of what it found and why it picked certain approaches.
Hypothesis generation
Based on what it sees in your data and what the literature says, the agent proposes testable hypotheses. These might be expected (confirming known biology) or unexpected (patterns worth investigating).
Treatment A upregulates inflammatory response genes
Analysis plan
Before running anything, you see the full plan: which methods, in what order, with what parameters. You can approve, modify, or ask questions. Nothing executes until you say so.
Multi-method analysis
The agent runs multiple methods on each step and compares results. If two approaches disagree, it flags the discrepancy. It can also adjust parameters automatically when initial runs suggest a better fit.
Human in the loop
At each checkpoint, you review results and decide what happens next. Ask follow-up questions, request different visualizations, override a decision. The agent explains its reasoning when you ask.
Export and share
Download the final report as PDF or HTML. Figures are formatted for publication. The code, environment, and data checksums are bundled together so anyone can reproduce the results. Share via link; no login required.