Request access
The Food Vector™ Platform

Most drug discovery starts in the lab. We start with food.

Food Vector™ transforms scientific literature, biological knowledge, and molecular data into testable candidates from food-derived molecules and peptides. Every candidate comes with a mechanistic hypothesis, supporting evidence, clearly stated uncertainty, and a path to validation.

Discover Food Vector See a Client Case Study →
Food Vector™ Platform

How it works.

Inside the Platform
One Connected Engine
Food Vector unifies your compounds, peptides, and research with its database, public science, structures, and pathways.

Everything is organized in one traceable molecular evidence graph, giving each candidate a connected path from source and structure to mechanism, score, uncertainty, and validation decision.

Your Knowledge

Compounds, peptides, products, assays, experiments, and proprietary research.

Food Vector Database

A growing search space of food-derived and food-inspired molecular representations.

Public Science

Literature, proteins, pathways, structures, ontologies, and biological evidence.

Food Vector™ Discovery
Two complementary discovery paths.

Use candidate-first analysis to map molecules and peptides to plausible targets and mechanisms, or biology-first discovery to identify candidates for a defined target or pathway.

Ligand-First

Give us a small molecule or peptide, and we map the human proteins and pathways it most likely acts on.

ligand → proteins and pathways

Receptor-First

Or give us a target, and we search ZERO STATE’s database of food-derived molecules and peptides for new binders.

receptor → food-derived candidates
Food Vector Physics
Food Vector Scoring Engine
Ranked on what matters.

Every compound is scored on a full developability profile — ADMET, permeability, stability, and other proven metrics — not a single binding number. The result is one composite rank you can compare fairly from one compound to the next.

Restoration
Safety
Developability
Selectivity

Food Vector weighs biological relevance, evidence, structural plausibility, safety, stability, and developability to show what should move forward.

The Reasoning Engine
The next experiment, not just an answer.
Grounded in the molecular knowledge graph

Every candidate is explained through the graph, with the exact structures, pathways, and sources behind each claim tracked and traceable. Candidates that need more evidence are flagged, and the most promising experiments to run next are surfaced.

How every candidate is explained
structure · 6QNO
assay · functional restorationRUN THIS
binding · Kd
pathway · restoration (hypothesised · in validation)
Worked examples
Quercetin · dietary polyphenol → target XADMET · permeability / tox flags
Food-derived peptide → target Xstability · protease + immunogenicity
⚠ off-target selectivity, evidence can’t yet support
Validate + Decide
Physics-based studies of binding and dynamics provide the best possible models of real biology.

Docking identifies plausible binding modes. Molecular dynamics tests whether those interactions remain stable over time. Free-energy calculations help distinguish durable candidates from attractive false positives.

01

Binding pose

Identify physically plausible interactions.

02

Dynamic stability

Test whether binding persists over time.

03

Confident decision

Separate durable candidates from false positives.

DECISION STANDARD

Advance the candidates that remain stable, selective, and physically plausible across methods.

Modalities Small molecule Peptide Protein target
Client Case Study

Global Bio Ingredients & Food Manufacturer
Unlocking the science behind natural therapeutics

Two natural compounds. Anecdotal reports of neurological benefit. No explanation for why they worked. Our mission was to find out — and to do it at a scale no wet lab could match.

2
Natural small molecules investigated for their neurological effects
~200
Peptides ranked to surface the strongest therapeutic candidates
~5,000
Human proteins screened by in-silico molecular docking
CHALLENGE 01 · NEUROLOGICAL

Why do these compounds affect the brain?

We docked both compounds against ~5,000 human proteins, then applied Gene Ontology enrichment analysis to reveal which pathways they preferentially engage — far beyond what chance would predict.

Key neurological pathways identified
High-affinity target proteins pinpointed
CHALLENGE 02 · GASTROINTESTINAL

What makes this peptide product heal ulcers?

A proprietary product of ~200 peptides, remarkably effective at healing stomach ulcers — but mechanistically unexplained. We applied the same framework to rank the pool and map the proteins behind the healing.

Top therapeutic candidates surfaced
Underpinning human proteins identified
The Discovery

The product doesn’t rely on a single healing mechanism — it engages multiple biological pathways simultaneously. A multi-target mode of action that likely explains its outsized clinical results.

Featured Use Case

Mechanism & Candidate Prioritization

Understand how a natural compound, peptide library, or existing product may work, identify the candidates most likely to drive the biological effect, and define what should be tested next.

Ranked candidate shortlist
Protein targets and pathways
Evidence, uncertainty, and rationale

Small-Molecule Discovery and Mechanism

Analyze known molecules, identify likely targets and pathways, rank candidates, and explain how they may work.

Peptide Discovery and Design

Analyze, rank, optimize, and design peptides for specific biological targets and pathways.

Biomarker Discovery

Connect molecular signatures to disease states and phenotypes.

Off-Target Assessment

Anticipate interactions and safety signals earlier in the pipeline.

Scientific Validation

Every conclusion is traceable to evidence.

Food Vector™ is designed for scientific scrutiny. Every output links to the literature, molecular structures, computational results, assays, assumptions, and reasoning steps behind it.

Evidence-linked outputs

Every claim connects to its supporting sources, data, structures, and calculations.

Reproducible analysis

Models, parameters, data versions, and computational steps are recorded for review and repeatability.

Expert-reviewed decisions

Domain experts can inspect, challenge, refine, or reject recommendations before candidates advance.

Auditable by design

Every recommendation maintains a traceable path from source evidence to candidate score, mechanistic hypothesis, and validation decision.

Illustrative · Evidence Trail insight_04821

A compound candidate modulates the target pathway via JAK/STAT inhibition.

Confidence
High
Literature
Multiple peer-reviewed studies associate the pathway with the target phenotype.
linked publications
Structure
The binding site is conserved across the protein family.
structure-referenced
Assay
Direction of effect is consistent with observed in vitro signal.
internal data
Every conclusion links back to the sources that produced it.