Literature Library

The Literature Library is your team’s shared collection of papers, under Data & Files. Upload PDFs and bioAF extracts the title, authors, DOI, and abstract automatically. Search across PubMed, bioRxiv, Europe PMC, and Semantic Scholar from one place. Comment on papers, track your own reading status, and associate papers with experiments or projects so they show up where your work lives.

On top of the library, AI Literature Review uses your active LLM provider to find papers you would otherwise miss, scoped to a specific experiment.

AI Literature Review

AI Literature Review asks the LLM to do the legwork of a manual literature search and triage. It is scoped to an Experiment and uses the same provider you configure for AI Review (bring your own key).

On demand. From an experiment, start an AI Literature Review. bioAF:

  1. Asks your active LLM provider to generate adjacent search queries from the experiment’s context.
  2. Runs every query across each enabled source and aggregates the candidates, de-duplicating by DOI.
  3. Asks the LLM to score each candidate for relevance (0.0 to 1.0) with a one-sentence rationale.
  4. Adds the top picks straight into your Library, associated with the source experiment and tagged with a note explaining why each paper was recommended.

The relevance bar (default 0.65) is configurable. Papers you have dismissed, and papers below your relevance threshold, are never re-recommended.

Automated. In Settings → Integrations → LLMs, turn on a daily, weekly, or monthly cadence. bioAF then runs AI Literature Review on its own for experiments that gained new samples or pipeline runs since their last automated review. You pick the first run’s date and time; it repeats on the cadence you choose. A configurable cap limits how many experiments run per cadence (oldest activity first; the rest roll to the next run). When an automated run adds papers, the affected users get an in-app notification.

Search across sources

Search four bibliographic sources in one place:

  • PubMed (NCBI), bioRxiv, Europe PMC, and Semantic Scholar

Each source is enabled per organization, with its own API key stored encrypted at rest. Search results stay outside your Library until you tick the ones you want and click Add to Library, so the Library only ever holds papers your team chose to keep. Duplicate DOIs are collapsed automatically, and inline HTML tags and entities in titles and abstracts are cleaned on ingestion.

Read, comment, and track status

  • In-app PDF reader: a paginated reader on each paper’s detail page, with Prev / Next, a page counter, and a Download link.
  • Automatic reading status: reaching the second page marks a paper Reading, and reaching the last page marks it Read. Status only ever moves forward, so it never undoes a status you set by hand.
  • Threaded comments per paper, so the team’s discussion lives with the work.
  • Independent filters for Active / Dismissed / Read / Unread / Reading, plus project and experiment association pickers. A paper can carry several status flags at once.
  • Associate and dismiss in bulk: multi-select papers and associate them with a project or experiment, or dismiss several at once. Dismissal is org-wide and reversible by an admin.

Export citations

Export any paper as BibTeX or RIS from its detail page. Bulk export covers every paper in the current filter or scope.

Literature in AI Review

Your curated library feeds back into AI Review. Abstracts and team comments on papers associated with an experiment can be bundled into the AI Review prompt; an admin controls which inputs are on per organization, project, or experiment, and can set a token budget so a large library does not blow up the review. AI Review of a pipeline run can then check results against the experiment’s associated literature and flag findings that are unexpected or that contradict prior work. When a paper’s full text is included, the assistant cites the page.

What is and is not sent

AI Literature Review follows the same trust boundary as AI Review:

  • To generate queries, bioAF sends the experiment’s context to your active provider.
  • To score candidates, it sends candidate paper titles and abstracts (public bibliographic metadata returned by the sources).
  • Search queries go to the external sources you enable (PubMed, bioRxiv, Europe PMC, Semantic Scholar).

Never sent: raw sequencing reads, the bytes of files on disk, or pipeline logs. As with AI Review, you bring your own API key and billing happens on your provider account.

Permissions

Adding papers, dismissing papers, and deleting papers from the Library are granted to the admin and comp_bio roles. Dismissals and deletions are org-wide and reversible by an admin. </content>