Literature and database search: papers, patents and sequence records

Biocompile’s literature and database search lets the assistant find papers, patents and public sequence records from a plain-language question. It stores the papers it selects with their full text, so you can read, highlight and query them, and every answer cites its sources.

Search the literature

The assistant picks the source that fits the question. These sources are available:

OpenAlex for structured scholarly metadata, with filters for year, type, open access and citations, and for retraction status.
Semantic Scholar for paper metadata, influential-citation signals and recommendations of related papers.
Patents, scientific documents and clinical trials, searched by concept with filters.
The web, for package documentation, standards, vendor documents and repositories.

Papers are looked up by DOI, PubMed ID, arXiv ID or the source’s own identifier. Agent Skills add guides for PubMed, bioRxiv and other literature sources.

Find open-access papers since 2022 on GH11 xylanase thermostability and recommend related work.

Store and question a paper

A stored paper is a paper whose full text Biocompile keeps on the project so it can be cited and queried. When you ask about a specific paper, the assistant stores it and asks targeted questions of the stored text, instead of answering from a search snippet.

Store the paper behind this DOI and tell me which expression host the authors used.

The assistant can also read a PDF or a web page it finds, and it keeps the text only when it stores the paper.

Read a stored paper

A stored paper opens as a record with tabs. The labels show a count where they apply.

Tab	What it shows
Overview	Why the paper is relevant, the abstract, key findings, methods, reagents, key results and conclusions
Full Paper	The full text with a table of contents
Figures	Extracted figures with their captions
Tables	Extracted tables
References	The paper’s structured references
Highlights	Your notes and the assistant’s, filtered by All, Mine or Agent
PDF	The original PDF

Key findings and methods are marked “AI-extracted from the full paper”. The record shows the paper’s licence, such as CC BY or CC BY-NC, as a tag.

To highlight, select text in the paper, choose Add note, write the note and choose Save. Biocompile shows “Highlight saved”. Edit a comment with Edit comment on its card, or remove the highlight with Delete highlight.

Query public databases

The assistant queries three public databases through managed tools.

Database	What the assistant can do
NCBI Datasets	Resolve gene symbols and IDs, search and summarize genomes, get taxonomy and virus reports, and return sequence download links
Ensembl	Annotate variants with the Variant Effect Predictor from HGVS notation, identifiers such as rsIDs, or genomic regions, and list assemblies and species
European Nucleotide Archive	Search studies by keyword, taxon or filters, read study details and linked publications, and list run files with checksums

For ENA files, the assistant generates a download script for you to review. It does not run the script.

Limitations

What is limited	The limit	What to do instead
Variant annotation in one batch	200 variants	Split larger lists into several requests
ENA study search	1,000 records per call	Narrow the search with filters
NCBI gene and genome download links	20 IDs or accessions per request	Ask in several requests
ENA downloads	A script is generated, not run	Review the script, then run it where you need the files
Full text of a paper	Needs the paper itself or an open-access link	Give the assistant the link or the file

Common questions

Which literature sources does Biocompile search?

Biocompile searches OpenAlex, Semantic Scholar, patent and clinical-trial documents, and the web. Agent Skills extend this to PubMed, arXiv, bioRxiv and medRxiv.

Does the assistant cite what it finds?

Yes. The assistant cites the papers and records it uses, and a stored paper keeps its full text on the project so you can check a claim against the source.

Can I search NCBI or Ensembl from a question?

Yes. Ask in plain language, for example for a gene by symbol and taxon or for the consequences of a variant. The assistant calls the NCBI Datasets, Ensembl or ENA tool that fits and reports the result.