Databases

It goes and looks it up.

A language model asked about a gene will tell you what it remembers. Inflexa opens the database instead, live, at the moment of the question, against the same sources a computational biologist would check by hand. Roughly thirty of them.

Install InflexaSee the tool definitions

Recall is not evidence.

When a model tells you a gene is associated with a disease, it is reporting a statistical impression formed during training. It may be right. It may be a confident sentence about an association that was retracted, or that never existed. You cannot tell from the sentence, and neither can the model.

So Inflexa does not ask the model to remember. It queries the source, records the query in the provenance, and cites what came back. A claim on a figure traces to a PMID or an accession, not to a recollection.

The sources

What Inflexa can reach

Sources marked with a key need an API credential of your own. Several are free to register for; a couple (DrugBank in particular) are licensed, and if you do not have access, that source is simply unavailable rather than silently substituted.

Genes, transcripts, and proteins

  • Ensembl

    Gene models, orthologs, annotation

  • UniProt

    Protein sequence and function

  • Human Protein Atlas

    Tissue and cell-type expression

  • NCBI

    Sequence, taxonomy, and gene records

Pathways and interactions

  • Reactome

    Curated pathways

  • KEGG

    Pathway maps

  • STRING

    Protein–protein interactions

  • MSigDB

    Gene sets for enrichment

Disease and genetics

  • Open Targets

    Target–disease association evidence

  • ClinVar

    Clinical variant interpretation

  • GWAS Catalog

    Published genome-wide associations

  • DisGeNET

    Gene–disease associations

Chemistry and drugs

  • ChEMBL

    Bioactivity, compounds, and targets

  • DGIdb

    Drug–gene interactions

  • PubChem

    Compound structures and properties

  • DrugBank

    Drug targets, pharmacology, and labels

Safety and pharmacovigilance

  • FAERS (openFDA)

    Adverse-event reports and disproportionality

  • PharmGKB

    Pharmacogenomics and CPIC evidence levels

  • EPA CompTox

    Chemical hazard and exposure data

Data and literature

  • GEO

    Public expression datasets

  • PubMed

    The published literature, queried live

  • PDB

    Protein structures

  • GitHub

    Method source and tool versions

This is not the complete list, and the complete list changes as tools are added. The authoritative version is the tool registry in the repository, which you can read.

Ask it something and watch it go look.

Every query is logged into the provenance record, so you can see exactly which source answered.

Install Inflexa