BioSkeptic
BioSkeptic is a red team for drug-discovery claims.
Core trick
Checks a target-disease hypothesis against about 21 public databases through eight falsifiable refutation mechanisms, returning cited concerns instead of a verdict.
Description
BioSkeptic is a red team for drug-discovery claims. You give it a target, a disease, and a direction; it checks that hypothesis against ~21 public biomedical databases (Open Targets, GTEx, the Human Protein Atlas, ChEMBL, ClinGen, PubMed, ClinicalTrials.gov, the FDA label, and more) and returns cited, grounded concerns — never a verdict.
The core is a set of refuting mechanisms drawn from the target-validation literature — each a specific, falsifiable question (Is the target expressed in the affected tissue? Do human genetics point the right therapeutic direction? Does the mouse knockout show anything relevant?). We catalogued ~18 and implemented 8, then benchmarked each against real approved indications and expert-refuted links, so the agent weighs a reliable flag more than a noisy one.
It's built for the scientist vetting a target before a program commits, and the computational biologist cleaning target–disease data before it trains a model. It ships as a web app, an MCP server, and a Python library, and can audit entire knowledge graphs in bulk. It gives you a grounded, checkable second opinion, early.
README excerpt
# BioSkeptic **Try it live: [bioskeptic.onrender.com](https://bioskeptic.onrender.com/)** BioSkeptic gives an AI better tools to check a drug–target–disease idea against real biological evidence — so its answer is something you can **verify**, not just something that sounds right. **▶ [Watch the 3-minute demo](https://www.youtube.com/watch?v=oYrEbjehWsg)** ## Introduction Most drugs fail, and they fail for all kinds of reasons — they don't work, they aren't safe enough, they aren't worth the cost. Much of that you only find out late, in the clinic. But one thing you *can* sanity-check early is whether a drug–target–disease idea even holds up against what we already know. People increasingly put that question to an AI, which is reasonable — but a language model answers from memory, and it sounds just as confident whether or not there's real evidence behind what it says. You get something plausible, with no easy way to check it. BioSkeptic gives the model better tools for that one question. It checks the idea against real biological data — is the target active where the disease is, do the genetics line up, has anything like it been tried — and returns the concerns it finds, **each with a link to its source**. It doesn't tell you yes or no; it just makes the answer checkable. The same tools work at scale, too: flagging the shaky links in a knowledge graph or dataset before a model is trained on them. It's one part of the picture — not a full assessment of a drug program —