KAIS Systems — molecular mechanism knowledge graph

Every edge carries the sentence it came from.

KAIS Systems reconstructs signed, directional molecular mechanisms across genes, proteins, metabolites, drugs and diseases — from over 100 million interactions extracted from the published literature. Any claim can be checked in one click.

Send a gene list How it works
One interaction, dissected
GeneTNF
upregulates expression positive · directed
GeneIL6

Stimulation with TNF-α produced a marked increase in IL-6 transcript levels in cultured synovial fibroblasts.

PMID 9284957 · Homo sapiens · synovial fibroblast · extraction confidence 0.94

Illustrative records. Every interaction in the knowledge base resolves to this structure — typed nodes, a signed directed relation, and the source text.

100M+Interactions
272Tissues resolved
8Entity types
14 yrsContinuous development

Method

Three properties that decide whether a mechanism is usable.

Association is cheap. What a reviewer, a regulator or a programme committee needs is direction, evidence, and the ability to get the same answer twice.

01

Direction and sign

We record that A upregulates B, not that A and B are associated. Relations are typed across expression, activity, modification, degradation, transport and binding — each one directed and signed. That distinction is what separates a mechanism from a correlation.

02

Provenance to the sentence

Every edge carries the PMID and the sentence the claim was extracted from, along with organism and cell line. Nothing has to be taken on trust, and nothing has to be traced back through a citation chain by hand.

03

Reproducibility

Knowledge base releases are frozen, dated and versioned. The same query against the same release returns the same result in three years — which is what a regulatory submission requires and what a generative system cannot offer.

Compare

Where we differ from the tools you already use.

Run the same input through all of them. We would rather you did.

CapabilityKAISIPAMetaCoreSTRING
Signed, directed relationsYesPartialPartialNo
Source sentence per edgeYesCitation onlyCitation onlyScore only
Genes and proteins as separate nodesYesNoNoNo
Tissue-resolved networks272 tissuesLimitedLimitedNo
Path-pattern queriesYesNoNoNo
Custom knowledge base buildsYesNoNoNo
Plant and agricultural coverageYesLimitedLimitedYes
Frozen, versioned releasesYesAnnual cycleAnnual cycleYes

Work with us

Three ways in, depending on how much you want to run yourself.

Engagement

Mechanism evidence package

Send us a gene, protein or metabolite list, or a compound and an endpoint. We return the reconstructed networks, hub and tissue analysis, enrichment statistics, and the complete evidence table with every supporting sentence. Four to six weeks, fixed price.

For teams with a deadline and no tooling

Build

Custom knowledge base

A knowledge base built for your compound class, crop, therapeutic area or endpoint set — using the same extraction pipeline, tuned to your domain ontology. Delivered as a versioned release you own and we maintain.

For programmes the general databases do not cover

Licence

Platform access

Run it yourself. Query wizard, network expansion, path-pattern templates, filtering and export, against frozen knowledge base releases. Per seat or site-wide, on our infrastructure or inside your firewall.

For groups doing this work continuously

Record

Published applications.

The method has been applied and peer-reviewed across human disease, agriculture, toxicology and extreme-environment physiology.

Infectious disease

SARS-CoV-2 metabolic dysregulation

Plasma metabolomics across COVID-19 patients and controls, with networks reconstructing how viral proteins regulate the perturbed pathways. Published in Scientific Reports.

Metabolic disease

Glucose variability in diabetes

Networks linking genes associated with both hyperglycaemia and hypoglycaemia, resolved across pancreatic, cardiovascular, adipose, muscle, renal and gastrointestinal tissue.

Space physiology

Cosmonaut proteome after long-duration flight

Urine and blood proteome analysis following ISS missions of 169 to 199 days, identifying cardiovascular markers linked to inflammation, coagulation and radiation response.

Agriculture and toxicology

Crop genetics and compound toxicity

A dedicated potato knowledge base spanning genetics, markers, breeding, pathogens and pests — and regulatory pathway reconstruction for chemical toxicity mechanisms.

Start here

Send us a gene list.

The fastest way to evaluate this is on your own data. Send a list of genes, proteins or metabolites and the question you are trying to answer. We will run it and walk you through what came back, with the evidence attached.

Email the team