How AtlasBio Works
A source-first literature search engine for biomedical research. Papers come first, the AI synopsis comes last, and every claim is anchored to a passage you can open.
We find the most relevant papers
When you ask a question, the system searches across every paper in the database and surfaces the ones most likely to answer it. Built specifically for biomedical literature, it understands the language of the field.
We pinpoint the exact passage
Inside those papers we identify the specific paragraph that most precisely answers your question. That paragraph is what you see attached to each result in the ranked list.
Synopsis last, not first
Only after you have the papers does the system write a short two-paragraph synopsis at the bottom of the page. It cites the ranked papers by number and cannot make a claim without a passage to back it.
Built for researchers, not chat sessions
Every design choice favors verifiable sources over slick conversation.
Literature Search
Ask in plain English. The retrieval reads inside each paper, so you do not need to guess at the right keywords. Works for broad questions (“mechanisms of central sensitization”) and narrow ones (“does HCN1 knockout reduce mechanical allodynia in SNI”).
Hypothesis Test
Propose a claim. The system pulls supporting and contradicting evidence in one pass and reports a structured verdict with confidence, identified gaps, and the strongest opposing papers. Good for stress-testing a draft before you commit it to a grant aim.
Matched Passages
Each ranked paper exposes the exact paragraph that earned its rank. Expand to read it, then open the DOI. No more chasing phantom citations.
Curated Metadata
Each database has domain-specific fields (species, injury model, anatomy, bone type, etc.) shown as inline badges so you can triage relevance without reading the abstract.
Save Your Ranked Sets
Registered users can save the full ranked paper list for any query. The saved set re-opens in the same source-first layout so you can come back to the actual evidence, not just an AI summary.
Show More Papers
If the top 10 are not what you wanted, ask for the next set. The system re-runs the query excluding what you already saw and tags the supplementary results so you know relevance may fall off.
Quick tutorial
Three steps to your first evidence set.
1. Pick a database
Go to Databases and choose the one closest to your question. Each database covers a different topic, so a pain question goes to a pain database, not the bone one. If you are unsure, pick the one that best matches your topic area.
2. Choose Literature search or Hypothesis test
Literature search is for open questions: “What is the role of microglia in neuropathic pain?” You get a ranked list of papers plus a short synopsis at the bottom. Hypothesis test is for claims you want to stress-test: “TRPV1 antagonism reduces mechanical allodynia after SNI.” You get supporting evidence, contradicting evidence, and a verdict.
3. Scan papers first, synopsis second
The results page is ordered by intent: numbered paper cards at the top with a one-line plain-English summary, then the short LLM synopsis at the bottom. Click View matched passage on any card to see the exact paragraph the ranker used. Open the DOI in one click.
Optional: save it for later
Create a free account to keep the full ranked paper set under your dashboard. From there you can re-run the same query for a fresh set of papers (the system excludes what you have already seen), or jump back into the saved sources.
Frequently asked questions
How is this different from Google Scholar or PubMed?
Google Scholar and PubMed rank papers by title, abstract, and citation graph keyword matching. AtlasBio reads inside the papers and ranks by the actual passage that answers your question. You can ask in full sentences without engineering keywords, and you see exactly which paragraph triggered each match.
How is this different from ChatGPT or a general LLM?
A general LLM generates an answer from training data and often fabricates citations. AtlasBio retrieves real papers first, shows you the source passages, and only then writes a short synopsis grounded in those passages. If a claim is not backed by a retrieved passage, the synopsis cannot make it.
Where do the papers come from?
Open-access biomedical literature, primarily from PubMed Central and PubMed. Each database is a curated set of papers built around a topic area (pain, bone, neurodegeneration, and so on).
How accurate are the citations?
Every citation links to a real paper with verified bibliographic metadata (authors, title, journal, year, DOI, PMID). The matched passage is the verbatim text from that paper, not a paraphrase. The synopsis is generated, but it cites by paper number from the ranked list above.
What is the difference between the two modes?
Literature search answers an open question with a ranked set of papers and a short synopsis. Hypothesis test evaluates a specific claim by actively pulling both supporting and contradicting evidence and weighing them in a structured verdict. Use Hypothesis test when you want to know whether a claim survives the literature.
How long does a query take?
30 to 90 seconds for most queries. The very first query against a database in a while can take a bit longer as the system gets ready. Subsequent queries on the same database are faster.
What if the top 10 papers are not what I wanted?
Click Show different papers at the bottom of the ranked list. The system re-runs your query while skipping what you already saw, and surfaces the next batch. A banner warns you that those results may be less relevant than the first set.
Can I query multiple databases at once?
Not for now. Each query targets a single database. If you need to cover more than one topic area, run the question against each database separately.
How are queries counted toward my limit?
Anonymous users get 2 queries per 7-day window. Registered users get 5 queries per 72 hours. Each submission counts as one query, including the Show different papers re-run and Hypothesis tests. Limits are global across databases.
What happens to my queries and saved sets?
Anonymous queries are not persisted to your account. Registered users can save individual query results to their dashboard. We log all queries for rate limiting and abuse detection, but saved sets stay private to your account and can be deleted at any time.