Not one answer. A population of hypotheses, scored and evolved.
Ask a question. Mendel finds the papers, breeds the ideas, and writes up the winner.
Watch competing hypotheses evolve.
Ask a research question and answer a few clarifying ones. Mendel retrieves up to 300 papers, writes a set of hypotheses, then forks and merges them over several rounds, scoring each on evidence, feasibility, impact and novelty. Watch the family tree grow, open any branch, and read the strongest as a full research proposal.
Which chromatin-remodelling dependencies are synthetic lethal with ARID1A loss in ovarian clear cell carcinoma?
Broad SWI/SNF inhibition
Target core subunits across clear cell carcinoma models.
ARID1B paralog dependency
Residual BAF complexes rely on ARID1B.
Generic checkpoint blockade
Not specific to the ARID1A-mutant lineage.
SMARCA4/SMARCA2 ATPase dependency
Loss of both ATPases collapses remaining remodelling.
ncBAF component BRD9
An orthogonal synthetic-lethal axis.
Combined ARID1B and SMARCA2 dependency in ARID1A-null cells
Merges two branches: test dual knockdown against single knockdowns in ARID1A-mutant and wild-type lines.
We audited our own AI scientist.
When Mendel revised a hypothesis, the step doing the revising was handed a different list of papers from the step that wrote the original. In 17 of 18 research questions the two lists shared no paper, and a word-overlap step filled most of the revised bibliography afterwards. Nothing errored, and the citations looked fine. We measured it and changed the revising step to read the same relevance-ranked papers.
Read the paper →The evidence your lab reasons from.
Save papers from search into shared workspaces. Each paper opens with its abstract, a note on why it's relevant, highlights, methods and key concepts. Ask Mendel questions across everything you've saved, and its answers cite the papers they draw on.
Residual BAF complexes in ARID1A-deficient tumours
Shows which subunits ARID1A-null cells still depend on.
Paralog dependencies across SWI/SNF-mutant cancers
CRISPR screen evidence for ARID1B and SMARCA2.
Chromatin accessibility in clear cell carcinoma models
Baseline profiles for ARID1A-mutant lines.
Residual BAF complexes in ARID1A-deficient tumours
Abstract
Loss of ARID1A leaves cells reliant on the remaining BAF complexes. We map which subunits those complexes require and test their loss across a panel of ARID1A-mutant and wild-type lines.
Why it's relevant
Directly tests the paralog dependencies your question is about.
Highlights
- ARID1B loss is selectively lethal
- Effect absent in wild-type lines
Methods
Key concepts
Type @ to cite · Enter to send
Search four databases at once.
arXiv, PubMed, Semantic Scholar and OpenAlex in one query, each result with a short note on why it matters to your question.
Shared workspaces with real roles.
Invite colleagues as viewers, editors or admins, and see what your team is saving as it happens.
Write it up, export to LaTeX.
Workbooks combine a block editor, AI drafting that asks for your perspective before it writes, scored feedback on research papers, and export to NeurIPS or IEEE templates.
Start your lab's workspace.
Search, save and write with your lab today. Explore is in private beta. Free during alpha.
Delete your account and everything in it at any time.