§ PLATFORM2026

An end‑to‑end discovery engine, built for translation.

Three operational layers — computational, experimental, and translational — held together by one biological data graph. Programs move forward continuously, not in monthly handovers.

01
COMPUTATIONAL BIOLOGY

Predictive models trained on real biology.

A foundation model trained on 1.2B annotated biological sequences predicts protein–compound interactions in silico — surfacing leads weeks ahead of physical assay.

Unlike vendor screening libraries, our graph is built from first‑principles biology: structural data, longitudinal patient cohorts, organoid readouts, and over a decade of curated literature. It updates nightly with every wet‑lab result that returns from the floor.

  • /01 Multi-modal target identification across genomics, transcriptomics and proteomics.
  • /02 Generative chemistry proposing 10⁵ optimisable analogues per scaffold, ranked for potency, ADMET, and synthesisability.
  • /03 Active-learning loop selects the next 1,000 compounds the wet lab should run each week.
FIG.01 — TARGET × COMPOUND GRAPHn=1.2B
Sequences
1.2B
Annotated, multi-modal
Compute
1.4PFLOPs
Internal cluster
In-silico → wet
94%
Concordance, audited Q1
02
AUTOMATED WET LAB

Robotic discovery on patient biology.

A 32,000 sq.ft robotic facility synthesises and tests thousands of candidates per week against patient‑derived organoid models — the highest‑fidelity in vitro assay system commercially available.

Every program runs on disease tissue, not transformed cell lines. Phenotypic readouts feed back into the computational graph within 36 hours of plate completion. Failure modes that would only emerge in late preclinical surface here, weeks earlier.

  • /01 Patient-derived organoid biobanks across 12 indications.
  • /02 Robotic synthesis floor with 24/7 throughput, 7,200 reactions per week.
  • /03 Multi-modal phenotypic imaging, mass-spec, and single-cell RNA readouts on every plate.
FIG.02 — ORGANOID ASSAY ARRAY252-well
Floor
32k sqft
Robotic wet lab
Reactions/wk
7.2k
24/7 synthesis
Hit→lead
14wk
Median, all programs
03
TRANSLATIONAL CLINICAL

From candidate to clinic, in‑house.

A biomarker‑led clinical team carries promising candidates through IND‑enabling studies and Phase 1 in partnership with academic medical centers across three continents.

Translational biomarkers are designed in alongside discovery, not bolted on at IND. Patient stratification hypotheses are pressure-tested against prospective cohorts before the first dose. Partners get a ready-to-run Phase 1 protocol, not a candidate molecule on a shelf.

  • /01 Prospective biomarker design integrated from program initiation.
  • /02 DMPK, tox, and CMC delivered via validated CRO network and internal preclinical lab.
  • /03 Investigator network across Karolinska, Memorial Sloan Kettering, and INSERM.
FIG.03 — PHASE 1 BIOMARKER RESPONSEn=84
Phase 1
4active
Programs in clinic
Centers
11sites
Across NA / EU / APAC
Patients enrolled
340+
Trailing 12 months

A program, end to end.

A typical engagement runs from target nomination through Phase 1 readout in 26–32 months — with continuous data sharing, no quarterly reveals.

WK 0–4

Target nomination

Patient multi-omics analysis to identify and validate druggable targets.

WK 4–18

Lead generation

Generative chemistry, virtual screen, in-silico ranking.

WK 18–32

Wet-lab validation

Robotic synthesis and patient organoid assays.

WK 32–80

IND-enabling

DMPK, tox, CMC, GLP packages and biomarker design.

WK 80+

Phase 1

Investigator network, biomarker-led patient stratification.

See the platform in action.