Discovery Engine  ·  Spring 2026 Modality-agnostic  ·  Indication-agnostic Brioche Biotech, Inc.

The molecular design layer for precision medicine. Patient or program context in. Optimized therapeutic out.

By construction
Modality-agnostic and indication-agnostic by design. Small molecules, biologics, peptides, ASOs, nucleic-acid therapies. Oncology, rare disease, autoimmune, metabolic, neurology, cardiovascular, infectious disease. The intent object takes any biological target in any patient or program context — the rest of the engine doesn't care.
00  /  How the engine expands

Three tiers. One platform. Same engine across the stack.

The Discovery Engine answers a target question across a full escalation ladder. Tier 1 is the default — re-rank the entire known therapeutic space against the target. Tier 2 designs minimal-modification analogs when the known space falls short. Tier 3 designs novel matter when no precedent fits. Same engine, escalating answer.
Tier 01  ·  Re-rank

The known space, ranked.

Score the entire known therapeutic space — approved drugs, clinical-stage compounds, literature analogs — against the target conformation and patient or program context. Days, not months.

Tier 02  ·  Analog

Minimal-modification analog.

When the known space falls short, propose minimal-modification analogs of the closest scaffold — predicted profile, synthesis route, IP positioning. Hand-off ready for experimental validation.

Tier 03  ·  Novel

Full novel design.

When no precedent fits, generate novel matter against the target — virtual screen, generative chemistry, predicted ADMET and synthesizability — fed into the same hand-off package downstream engagements consume.

The Tier the engagement runs at depends on the question, not the buyer. A program team with a hit in hand starts at Tier 2. A precision-oncology team with a refractory mutation starts at Tier 1 and escalates as the known space exhausts. The engine moves between Tiers without rebuild.
01  /  Engagement formats

Four formats, matched to the discovery question, not the project plan.

Most discovery teams come to us with one bottleneck — an unvalidated target, a thin hit list, a stalled lead series, an asset they're being asked to in-license. The format below is matched to the shape of the answer, not the size of the program. Scope, deliverables, and hand-off are agreed in writing before kickoff.
Format I  ·  2026.D
Target Validation Sprint
Timeline2–4 weeks
OutputValidation dossier · ~70pp
AudienceCSO · VP Discovery · Board
Hand-offGo / no-go memo

A putative target arrives with a thesis. We return a validation package — druggability, disease linkage, failure history, competitive set, and IP terrain — structured so the program can decide to commit, refocus, or kill before chemistry spend begins.

What you receive
  • Druggability assessment — structural tractability (PDB coverage, pocket analysis, allosteric site survey) and chemical tractability (precedent ligand classes, hit rates in disclosed screens)
  • Disease-target linkage strength — human genetics (GWAS Catalog, OpenTargets), functional perturbation evidence, clinical biomarker correlations, model-system reproducibility
  • Failure analysis — programs that have pursued the target, terminated stages, public reasons for failure where disclosed, inferred reasons where not
  • Competitive landscape — active and historical programs across ClinicalTrials.gov, EU CTR, JapicCTI, conference disclosures, and partnered-asset news
  • IP and freedom-to-operate — composition-of-matter, method-of-use, screening-method coverage from USPTO PatentsView, Espacenet, and Google Patents
  • Go / no-go memo — one-page recommendation with the three risks the program should price into a chemistry decision
Best fit when

A discovery team has a target proposed by a biology group, an academic collaborator, or an in-licensing pitch — and the chemistry investment is large enough that the team wants an outside, structured validation read before the chemistry budget moves.

A question we'd answer

The biology group is pushing this kinase as our next program. Before we put a chemistry team on it for nine months, is the target druggable, is the disease linkage real, and are we second-, fifth-, or fifteenth-in-class on the public record?

Format II  ·  2026.E
Hit Identification Campaign
Timeline4–8 weeks
OutputHit package · SMILES + profiles
AudienceMed-chem lead · VP Discovery
Hand-offCRO sourcing brief

A validated target needs hit matter. We run a computational hit-finding campaign — virtual screening, generative chemistry, and pharmacophore-driven similarity searches — and return a triaged hit list ready for synthesis and experimental validation through your CRO of choice.

What you receive
  • Pharmacophore models built from disclosed actives in ChEMBL and the patent literature, with binding-mode hypotheses tied to public PDB structures
  • Virtual screening campaign across enumerable libraries (ZINC22, ENAMINE REAL, in-house collections if provided), with docking and pose validation
  • Generative chemistry runs targeted at scaffold diversity, with predicted profiles for each suggestion and synthetic accessibility scoring
  • Hit triage — predicted potency band, selectivity flags against off-target panels, ADMET red flags (hERG, CYP liability, PAINS, reactivity)
  • Synthesis prioritization — ranked list with predicted profile, sourcing route, and supplier availability where catalogue compounds exist
  • CRO hand-off package — SMILES, predicted profile sheet, suggested assay panel, and a sourcing brief written for synthesis and in-vitro pharmacology CROs
Best fit when

A program has a validated target but no internal med-chem capacity, a hit list that's too thin, or a screening campaign that returned scaffolds the team wants to expand computationally before committing more screening spend.

A question we'd answer

Our DEL screen returned three scaffolds with sub-micromolar binding but ugly ADMET predictions. Can you expand each scaffold computationally, triage to a fifty-compound short-list with cleaner predicted profiles, and prepare the sourcing brief so our CRO can synthesise next month?

Format III  ·  2026.F
Lead Optimization Sprint
Timeline4–6 weeks
OutputSAR memo · analog roadmap
AudienceMed-chem lead · CSO
Hand-offSynthesis priority list

A scaffold is in hand and the program is grinding through SAR. We deliver a structured optimisation plan — structure-activity analysis from existing data, computational suggestions across the four developability axes, and a ranked analog list to test next.

What you receive
  • SAR analysis from the program's own assay data (under NDA), supplemented by public-record analogues where the chemotype is disclosed
  • Computational suggestions targeted at the program's actual gating constraint — potency, selectivity, ADMET, or PK — rather than a generic "improve everything" set
  • Predicted profile of each proposed analog: potency estimate with confidence band, selectivity flags, ADMET prediction (solubility, permeability, hERG, CYP, hepatotoxicity), and PK proxy
  • Free-energy perturbation or relative-binding-affinity calculations on the highest-value analogs where the structural basis supports it
  • Synthesis prioritization — ranked list with predicted gain on the gating constraint, predicted loss on adjacent properties, and synthetic accessibility
  • SAR memo summarising what the data implies about the binding mode, what it doesn't, and which assays would resolve the open questions
Best fit when

A med-chem program is mid-optimisation, the easy moves have been made, and the team needs a structured outside read before committing the next twelve weeks of synthesis to a particular vector — or wants a competing computational hypothesis tested against the in-house plan.

A question we'd answer

We've made forty analogs and we're stuck at high single-digit nanomolar potency with a CYP3A4 problem we can't engineer out. Show us where the binding mode actually constrains us, propose ten analogs that resolve the CYP issue without giving back potency, and tell us which three to make first.

Format IV  ·  2026.G
Asset Diligence
Timeline2–3 weeks
OutputDiligence dossier · risk register
AudienceBD lead · IC · Diligence team
Hand-offIndependent risk read

A discovery asset is on the table for in-licensing, partnership, or acquisition. We deliver an independent computational re-read of the asset's claimed activity, target rationale, developability profile, and competitive position — written to be defensible in a deal room.

What you receive
  • Computational re-validation of claimed activity — binding-mode hypothesis check, predicted vs. claimed potency, selectivity panel re-scoring against public off-target sets
  • Independent target rationale check — disease linkage, competitive density, failure precedent, druggability re-read on the asset's own assumptions
  • Predicted developability profile — ADMET, PK proxy, CMC red flags from the disclosed structure where the deal package permits
  • Competitive and IP terrain — who else has filed in the chemotype, what the freedom-to-operate looks like, and where the seller's IP narrative is thinner than the slide deck implies
  • Risk register — ranked, quantified where the evidence supports it, mapped to the structure of the deal under negotiation
  • Independent confidence statement — one paragraph per dimension, with the evidence base and our band on each conclusion
Best fit when

A BD team or investment committee is being asked to take a position on someone else's discovery asset, and the seller's deck is the only computational read on the table. Brioche provides the second read — structured, sourced, and independent of the deal counterparty.

A question we'd answer

This seed-stage company is pitching us a pre-IND small molecule with a clean safety pharm package and aggressive potency claims. Re-validate the chemistry independently, tell us where the IP is thin, and price the three risks we'd carry if we in-licensed the asset at the proposed terms.

Optional  /  Co-development
Co-development partnership
For programs where the alignment of incentives makes more sense than a fixed-fee engagement, we will discuss equity- or milestone-based co-development on a case-by-case basis. The default is fee-for-service. Co-development is a deliberate choice on both sides, not a discount mechanism, and we are conservative about taking these on.
Illustrative
02  /  Specimen excerpt

What a deliverable actually looks like.

Below is an illustrative excerpt from a Target Validation Sprint — a druggability and disease-linkage section, with the citation framework intact. Sources cited are real public databases (OpenTargets, ChEMBL, PDB, GWAS Catalog) but the specific target, sponsor, and read have been generalised. This is the editorial standard every Discovery Engine engagement holds to.
DISC-VAL · Sprint 2026.04 · Section §2.1

Druggability & Linkage Read: a representative kinase target in fibrosis

Prepared for Sponsor (Redacted) Issued Apr 2026 Pages 11 / 68
Illustrative specimen — not a real engagement Brioche Biotech · Confidential to Sponsor
§ 2.1  /  Druggability and disease linkage

The target is structurally tractable. The disease linkage is the open question.

The proposed target is a serine/threonine kinase with high-resolution apo and ligand-bound structures deposited in the Protein Data Bank.[1] The OpenTargets association score for the indication of interest is mid-band, driven primarily by transcriptomic evidence rather than human genetics or rare-variant signal.[2] The Sponsor's working assumption, as captured in the kickoff brief, is that the structural tractability and the OpenTargets score together carry the validation burden. That assumption is partially, but not fully, supported by the evidence base.

On structural tractability, the read is unambiguous. The kinase has an ATP-competitive pocket with the canonical hinge geometry, multiple disclosed type-I and type-II inhibitor classes against orthologue or paralogue proteins in ChEMBL,[3] and at least one resolved allosteric site visible in the apo structures.[1] Hit rates for the chemotype family are in the expected band for kinase virtual screens; the structural argument carries.

For a discovery decision of this size, the load-bearing question is not "is the target druggable" — it is "is the disease linkage strong enough to spend chemistry on a five-year clinical bet." Those are different questions.

On disease linkage, the picture is thinner. The GWAS Catalog returns no genome-wide-significant association between coding variants in this kinase and the indication of interest at p < 5×10⁻⁸.[4] The functional evidence in OpenTargets is dominated by bulk-tissue transcriptomic upregulation in disease vs. control comparisons — informative, but not causal. The mouse model evidence is mixed: two independent perturbation studies show partial rescue, one shows no effect, and the heterogeneity has not been resolved in the published record.

Two specific risks follow. First, causality: transcriptomic association does not establish that pharmacological inhibition will reverse the phenotype, and the mouse model heterogeneity is consistent with the kinase being a marker rather than a driver in a meaningful patient subset. Second, patient selection: if the underlying biology is heterogeneous, a Phase-2 readout against an unstratified population is at material risk of failure even with a clean potency and ADMET profile.

Our recommendation, captured in §2.4, is to gate the chemistry decision on a single human-tissue functional experiment that the program can run within the validation window, rather than treating the OpenTargets score as the validation. The competitive-density section (§3) makes the case that this resolution is also a moat-creating exercise, not just a risk-management one.

Validation dimension
Source
Strength
Read
Risk band
Structural tractability ATP-pocket geometry, allosteric site
PDB · multiple
High-resolution
Tractable
Low
Chemical tractability Disclosed ligand classes
ChEMBL
Multiple chemotypes
Tractable
Low
Genetic linkage Common & rare variant
GWAS Catalog
Below threshold
Weak signal
Elevated
Functional evidence Perturbation studies
OpenTargets · literature
Mixed across models
Heterogeneous
Elevated
Translational signal Patient stratification
Disclosed clinical biomarker work
Limited
Inferred
Elevated
Brioche Biotech, Inc. · Discovery Engine Document control · DISC-VAL-2026.04 · Rev A
03  /  How an engagement runs

A scoped engagement, delivered on a calendar.

Most discovery engagements fail at the brief, not at the chemistry. We hold a long scoping conversation, write the brief back to you in plain English, and don't begin work until you've signed off on the question, the deliverable, and the hand-off shape. The calendar below is for a Target Validation Sprint — the longer formats follow the same shape, scaled out across additional weeks.
Week 0 · Pre-engagement
i.
Scope
A 90-minute call to articulate the discovery question. We send back a one-page brief with the question, the deliverable shape, the data sources we expect to bring, the hand-off (CRO brief, synthesis list, go/no-go memo), and what's out of scope. No work begins until the brief is signed and the NDA is executed.
Week 1 · Kickoff
ii.
Kickoff
A working session with the analyst pair on the engagement — a senior computational chemist and a discovery analyst. Source list reviewed, pre-existing data ingested under NDA where shared, open questions surfaced early so they don't surface as surprises in the deliverable. Calendar locked.
Weeks 2–3 · Production
iii.
Delivery
Computational runs, structured analysis, internal review against the citation framework, and an editorial pass. Mid-engagement check-in at the halfway mark when the brief allows interim findings. The dossier is delivered as a bound PDF with hyperlinked citations and, where applicable, machine-readable hand-off files.
Week 4 · Debrief
iv.
Debrief
A 90-minute debrief call: walk through findings, the risk register, and follow-up vectors. Up to two written follow-up memos within 30 days are included. Where the engagement produces a CRO hand-off package, we are available on the kickoff call with the synthesis or in-vitro CRO at no additional fee.
04  /  Wet-lab partners

We are computational. The bench work runs through CROs.

Honesty matters here. Brioche is a research practice in computational discovery, not a wet-lab operation. Synthesis, in-vitro pharmacology, ADME profiling, in-vivo studies, and analytical characterisation are coordinated through CRO partners — either yours or, where you'd prefer, partners we've worked with before. We do not run a lab and we do not pretend to.

Synthesis
Custom synthesis & scaffold expansion
Synthesis CROs of your choosing. Where you don't have a preference, we will recommend operators we've coordinated with on prior engagements and prepare the sourcing brief in their preferred format. Compound shipment, QC, and cost-of-goods accounting stay with the CRO contract; we do not handle materials or invoice for them.
In-vitro pharmacology
Biochemical & cellular assays
Target-engagement, functional, selectivity, and cellular potency assays run through your in-vitro CRO. The Discovery Engine deliverable specifies the assay panel we recommend, the order of priority, and the decision rule attached to each readout — so the assay vendor can scope and quote against a written brief rather than a verbal one.
ADME & safety pharm
Developability profiling
Standard ADME and safety-pharmacology panels — solubility, permeability, microsomal stability, hERG, CYP inhibition, plasma protein binding — coordinated through the CRO of your choice. Where the program is pre-CRO, we will scope a panel sized to the engagement and the decision in front of you.
In-vivo
PK/PD & efficacy studies
In-vivo PK, tolerability, and efficacy studies are CRO work. We help size the study, articulate the powering assumptions, and integrate the readout back into the computational model — we do not run the colony, we do not interpret the histopathology, and we do not own the IACUC submission.
Note on naming
No named partner logos
We do not list specific CRO logos on this page. Where we make a recommendation, it is informed by the engagement, the chemotype, the geography, and the program's existing relationships — not by a partnership we are paid to feature. If a CRO is genuinely the right operator for your program, we'll say so on the call.
05  /  Compared to alternatives

Where we fit, honestly.

Buyers comparing the Discovery Engine are usually weighing three other options — a traditional discovery CRO, an in-house computational team, or a software platform — or considering not doing the work at all and accepting the risk. Below is how we read those tradeoffs. We've also listed where we fall short, because the only thing more useful than knowing what a partner is good at is knowing what they aren't.
Option
Where it wins
Where it strains
Where Brioche fits
Traditional discovery CRO full-service contract research
Integrated chemistry and biology under one roof, FTE-priced, project-managed end-to-end.
Slow on the analysis layer, expensive against unbounded scope, fragmented across functions, and rarely opinionated on go/no-go calls because the incentive is to keep the FTEs running.
We're the structured, opinionated computational layer on top of a CRO contract — or before one is signed, when the question is whether to start.
In-house discovery team med-chem & comp-chem hires
Full institutional context, direct ownership, embedded judgement, ongoing availability across multiple programs.
Capacity-limited — a four-person comp-chem group cannot run six programs in parallel, and the marginal program waits. Expensive to build, slow to ramp, hard to flex.
We're the elastic capacity for a specific program, on a specific question, without taking permanent headcount.
Software platform SaaS modeling tools
Self-serve, standardised, runs on the team's own data on the team's own schedule.
No service layer. The platform produces outputs; the team still has to interpret them, integrate across functions, and make the decision — which is the work that was actually expensive.
We bring the service layer — the integration across data sources, the editorial pass, the recommendation, and the hand-off to wet-lab.
Doing nothing accept the risk, push forward
Zero direct cost; preserves the program timeline against scoping delays.
The cost of an unvalidated target reaching IND-enabling spend is the dominant failure mode in early discovery. The savings on validation are the smallest line item in a program that fails at Phase 2.
A Target Validation Sprint runs in two to four weeks against a fixed scope. The asymmetry against a multi-month CRO program is the entire argument.

Where we fall short.

We are not a wet-lab.
If the engagement is "synthesise these compounds and run the assay panel," we cannot do that. We can scope it, write the CRO brief, and integrate the readout back into the model — but the bench work runs through your CROs, not through us.
Targets that require novel modality work.
Our practice is strongest on small molecules and small-molecule-adjacent modalities. PROTAC and molecular-glue work is in scope; antibody discovery, cell therapy, and most novel-modality targets are not, and we will tell you so on the scoping call.
Programs that need the answer in a week.
Our shortest credible cycle on a Target Validation Sprint is two weeks. Faster than that, the source discipline collapses and the deliverable becomes a literature summary — which is not what we sell.
Engagements that are political, not analytical.
If the deliverable's job is to support a discovery decision that's already been made, we are a poor fit. Our reports are written to be useful, which sometimes means uncomfortable for whoever championed the program.
06  /  A note on method

Bound to the public discovery record.

Every Discovery Engine engagement holds to the same evidentiary discipline as Intelligence Studio — the same one a med-chem reviewer, a regulator, or an investment committee would expect. The five notes below describe the discipline and where it ends.

01.
Public sources, cited inline.
Every claim of fact is bound to a specific record — a PDB entry, a UniProt accession, a ChEMBL assay, an OpenTargets association, a USPTO application, a ClinicalTrials.gov record, a peer-reviewed publication. Marginalia carry the identifier. Inferences are flagged as such, with a confidence band, never as fact.
PDB · UniProt · ChEMBL · OpenTargets · GWAS Catalog
02.
Computational discipline, versioned.
Docking grids, screening libraries, generative models, and free-energy protocols are versioned and recorded with the deliverable. The same engagement, run again on the same evidence base, returns the same answer. Methodological choices are documented — not buried in a footnote.
ZINC22 · ENAMINE REAL · method registry
03.
Expert review, before delivery.
No deliverable leaves the practice without a second-analyst review against the citation framework and a senior med-chem editorial pass. The standard is not "we found something interesting" — it's "we'd defend this in front of a med-chem committee."
04.
No fabricated activity, by policy.
We do not present predicted potencies as measured. Where the deliverable carries a number, the source is named — computational prediction with a confidence band, public assay record, or sponsor-confidential data integrated under NDA. We do not concede the language of "validated" to a predicted profile.
05.
Audit trail, retained.
For each engagement, we retain the source pull list, the analyst working notes, the computational run logs, and the editorial review record for seven years. If a finding is ever questioned — in due diligence, in a partner negotiation, in a regulatory filing — the underlying record is reproducible.
07  /  Bring the question

Bring us the discovery question. We'll write it back to you.

New engagements begin with a 90-minute scoping call. We don't ask you to fill in a form — we ask you to describe the program, the bottleneck, and the decision in front of you. We send back a one-page brief with the question framed in writing, the deliverable shape, and the hand-off path.

Office
Brioche Biotech, Inc. · Delaware
Engagements
Currently accepting briefs for Q3 2026.
Confidentiality
Mutual NDA executed before scoping call.