How Motif Works

Ask a question, get structured biomarker evidence, and verify everything with source-linked citations.

Ask a research question

Enter your question in plain English, including disease context, biomarker classes, or therapeutic focus.

We extract biomarker relationships from the literature

Motif retrieves relevant papers, extracts biomarker-disease-drug relationships, and validates them against trusted databases.

Get structured, cited biomarker insights

Explore an interactive knowledge graph with PMIDs and evidence links so you can validate every claim.

Pipeline overview

A simple view of how a query becomes structured and cited output.

Question

Literature retrieval

Extraction & cross-referencing

Structured result graph

Example query

What biomarkers are linked to survival in KRAS-mutant pancreatic cancer?

Motif retrieves relevant literature, extracts biomarker-outcome relationships, validates them against clinical and genomic databases, and returns cited associations you can open and verify in the knowledge graph.

Why researchers trust this output

Source-linked at every step
Cross-referenced across 50+ databases
No black box—grounded in published literature

What you get

Structured biomarker associations
Evidence indicators to prioritize what to review first
Knowledge graph outputs for faster hypothesis generation
Literature review outputs: cited narratives, tables, and exports for grants and publications

Data sources

Motif cross-references evidence against trusted scientific and clinical resources:

PubMedClinVarClinicalTrials.govUniProtReactomeOpen Targets

+50 additional sources in genomics, pathways, immunology, and ontologies.

Technical details and limitations

Need the full methodology details? Expand each section below.

Methodology detailsExpand

AI extraction approach

Motif uses language models to extract biomarker-disease-drug relationships from published literature. We do not generate novel claims; we organize and cross-reference existing research with citations for verification.

Search coverage strategy

Precision-focused retrieval for targeted matching
Recall-focused retrieval to broaden evidence discovery
Terminology handling for alternate scientific naming

Biomarker and relationship coverage

Motif recognizes 60+ biomarker types and multiple relationship classes, including:

Genes and variants

Proteins and antibodies

Metabolites and lipids

Epigenetics and ncRNA

Immune and gene signatures

Pathways and phenotypes

Predictive and diagnostic

Prognostic and monitoring

Drug sensitivity and resistance

Additional outputs

Patent intelligence: cross-reference biomarker and therapeutic claims against patent and regulatory sources where integrated.
Meta-analysis and synthesis: aggregated evidence views where comparable study data supports comparison.

Full database list

Literature

PubMed - Biomedical literature

PubMed Central - Full-text archive

Europe PMC - European research

Clinical

CIViC - Clinical variant interpretations

ClinVar - Genetic variant significance

ClinicalTrials.gov - Clinical studies

DGIdb - Drug-gene interactions

MedGen - Medical genetics

Open Targets - Drug targets

PharmGKB - Drug-gene relationships

Genes

HGNC - Gene nomenclature

NCBI Gene - Gene information

Ensembl - Genome annotation

MGI - Mouse genetics

RGD - Rat genetics

Genomic Variants

gnomAD - Population genetics

dbSNP - Genetic variants

cBioPortal - Cancer genomics

Proteins

UniProt - Protein sequences

Human Protein Atlas - Protein localization

GtoPdb - Pharmacology targets

Pathways

Gene Ontology - Gene functions

Reactome - Biological pathways

STRING-DB - Protein interactions

Immunological

ImmPort - Immunology data

InnateDB - Innate immunity

IEDB - Immune epitopes

Antibody Registry - Antibody data

IPD-IMGT/HLA - HLA sequences

IPD-MHC - MHC sequences

VDJdb - TCR sequences

Metabolites

ChEBI - Chemical entities

LIPID MAPS - Lipid structures

Glycans

GlyTouCan - Glycan structures

GlyGen - Glycoscience data

GlycoEpitope - Glycan epitopes

Ontologies

MONDO - Disease ontology

UBERON - Anatomical ontology

HPO - Human phenotypes

Chemical

ChEMBL - Bioactivity data

Regulatory

FDA - Drug approvals

ATC - Drug classification

RNA

RNAcentral - Non-coding RNA

Epigenetic

Ensembl Regulatory - Regulatory features

PTM

iPTMnet - Post-translational modifications

Gene Signatures

MSigDB - Molecular signatures

Cellular

Cell Ontology - Cell types

Taxonomy

NCBI Taxonomy - Organism taxonomy

NCBI Virus - Viral sequences

Processing-time context

End-to-end runtime varies by query scope, evidence volume, and system load. Internal testing indicates per-article extraction around three minutes under representative conditions. This is one stage of the pipeline, not total run time.

Limitations and responsible useExpand

Motif is a research acceleration tool and should be used with expert review.

Coverage gaps may exist for preprints, very new papers, or under-indexed topics.
AI extraction can miss context or misclassify relationships in edge cases.
External database updates may lag behind newly published findings.
Evidence indicators reflect literature support, not clinical validity predictions.
Researchers should verify key findings by reviewing the original source papers.

Questions about methodology?

If you want deeper technical clarification or want to report an issue, contact us.