An idea for a drug should be enough to start

The belief

Whether an idea gets tested should depend on the idea.

Today it depends on the balance sheet behind it. A discovery programme takes five vendors and six figures, so the filter on which therapeutic ideas get tested is capital rather than scientific merit. We think that is the wrong filter, and we are building the route around it.

The filter is capital

A biotech that cannot commit six figures and manage five vendors cannot run discovery at all. The idea is never tested, and nobody finds out whether it was right.

The ideas sit elsewhere

Deep target biology usually sits with a small team or an academic spinout. The insight and the means to act on it are held by different people.

So we built the route

One integrated pipeline, priced so an idea is enough to begin. That is the company. Everything else is how we do it and what it costs.

What it costs today

Good ideas die for want of a discovery budget.

01

One programme takes five vendors

Antigen expression, library construction, screening, affinity measurement and developability are separate contracts, quotes and calendars. The customer is left to integrate them.

02

Two prices, neither of which works

A bare phage campaign runs about £8k and returns unranked binders. Full outsourced development runs £25k to £250k. A small biotech either under-scopes or cannot approve the spend.

03

Failure arrives after the money is gone

Binders that clear a binding assay still fail on aggregation, thermal stability and immunogenicity, at the bench and in the clinic, once the budget is committed.

The demand is already moving. Roughly half of antibody discovery is outsourced today, and the outsourced share is growing at 17.3 per cent a year, faster than in-house discovery.

Why now

The belief only became buildable last year.

We could not have built this two years ago. Two independent groups published working de novo antibody design within months of each other last year, and the methods are open, so the expensive step stopped being the hard one.

<0.1%The historical success rate for de novo antibody designChai-2

16%Overall hit rate, 20.0 per cent for VHH and 13.7 per cent for scFv, measured by BLIChai-2

4–22%Wet lab success across four targets, best nanobody at 140nMGerminal, Arc Institute

200 / 1,400Approved antibody therapeutics and candidates in clinical studies. The market we scale intoThe Antibody Society

Where the constraint went

The industry is still racing on generation while the two steps either side of it decide outcomes: choosing which site on a target to attack, and carrying a candidate through validation to something manufacturable. Owning both is what an integrated pipeline is for, and it is buildable now precisely because generation stopped being the hard part.

The pipeline

One pipeline. Idea in, candidate out.

In house, computational Partner laboratories today, in house from year three

Target and epitopeStep 01

AI generationStep 02

Selection layerStep 03

Ranked shortlistStep 04

Wet lab validationStep 05

Delivered candidateStep 06

One contract, one calendar

We own the whole chain: target analysis, design, selection, wet lab validation and delivery. The customer manages one relationship instead of five and carries no integration risk.

Ranked before the bench

Every candidate is scored on how designable its target site is and how manufacturable the molecule is, before any wet lab money moves. Nothing reaches a laboratory unranked.

One format first, then all of them

We start in VHH nanobodies because they are the fastest format to validate. The selection layer is scaffold agnostic, so scFv, Fab, IgG and bispecifics follow without rebuilding it.

The mechanism

Where you aim decides everything.

A high probability of success is not a claim about our generator. It comes from choosing the target site before committing to it. The same method, on the same target, swings by two orders of magnitude depending on that choice.

0.4% Whole target, sites picked at random. 10 hits from 2,317 designs mBER, Manifold Bio

7.5% Same target, best-performing site. 6 hits from 80 designs mBER, Manifold Bio

38% Favourable sites under filtering. 8 hits from 21 designs mBER, Manifold Bio

A hundredfold difference from one decision, on one target, using one method. Our screening layer makes that decision. It is a property of the target surface rather than of the binder, so the same advantage carries into every format we add.

Two decisions, both ours

Before and after generation

Before generation we rank every surface patch on the target by how designable it is, so effort only goes where binders can be found.

After generation we rank candidates on manufacturability against clinical-stage distributions, so what we deliver can be made. Four axes carry that ranking.

AxisEpitope designability

AxisThermal stability

AxisAggregation propensity

AxisHumanness and immunogenicity

Thresholds are set from the observed distribution of clinical-stage molecules. Applying that method independently reproduced the Oxford profiler's published bands exactly, which is our evidence the ranking is sound.

Why it stays ours

Generation is commodity. Selection is not.

Generation is improving in the open and nobody holds an advantage in it for long. We treat it as an interchangeable input and expect to swap the engine as better ones arrive.

Selection does not work that way. Every campaign we run adds paired design decisions and measured outcomes on a protein surface nobody else has screened, and that accumulates into something a competitor cannot buy.

Developable leads first. Fewer wasted cycles. A lower cost per candidate that works.

What we sell

Four products. Pricing rises as we take more of the chain.

Screening report

Epitope designability and developability ranking. No wet lab.

£15–30k

Now

Managed campaign, partner laboratory

The full chain, with wet lab validation run through our CRO partners.

£80–150k

Now

Managed campaign, own laboratory

The same deliverable with the laboratory in house.

£120–200k

Year three

Retained multi-target programme

A standing discovery function for one customer.

£400k–1m

Year four

Why us

Discovery CROs return unranked binders and leave the developability risk and the site gamble with you. AI-native design companies build their own drugs on closed platforms, so their incentive is to keep good targets rather than hand them back. Open academic tools are free, and you supply the GPU, the expert, the CRO and the risk.

We are the only route that owns target analysis through to validated molecule under one contract.

Where we are

Built, deployed and calibrated.

The platform is operational and not a prototype. It runs end to end today, which means the work in front of us is evidence rather than engineering.

Pipeline built and deployed

Cloud deployed on AWS with A100 compute. Runs end to end from target input through generation, selection and ranked shortlist to construct design and wet lab handoff.

Selection layer calibrated

Calibrated against clinical-stage distributions and a public dataset of 496 nanobodies, and independently reproduced the Oxford nanobody profiler's published thresholds.

Adversarially tested

Both the developability screen and the epitope tool have been through full audit and fix cycles, with regression tests verified against pre-fix code.

Working demonstration

An end-to-end walkthrough on real PDL1 data, from target input to a ranked, construct-ready shortlist.

Wet lab validation In progress

The first campaigns are being scoped with our CRO partners, structured so they return developability measurements alongside binding data. Every figure quoted on this page so far is published work by the groups named beside it. Ours will be added here as the campaigns report.

Formats

Five formats running, from single domain to full IgG.

We run VHH, VH, scFv, Fab and IgG today. Every step of the chain is scaffold agnostic, so a format is a template rather than a rebuild, and the same target analysis, generation and selection layer carries across all five. A customer picks the format the programme needs rather than the one the platform happens to support.

Wet lab validation starts in VHH, because it folds reliably, expresses cheaply and returns binding data in weeks rather than quarters, which makes it the fastest honest way to put numbers against the pipeline. The larger formats run alongside it, so the market we scale into is the whole antibody market and not the single-domain corner of it.

VHHRunning

VHRunning

scFvRunning

FabRunning

IgGRunning

BispecificPlanned

Who we are

ML first, closing the gaps deliberately.

Abhi

Technical Co-founder

Mathematics and data science, LSE. Built the discovery pipeline and the selection layer. Leads engineering and the technical roadmap.

Alexander

Scientific Co-founder

Neuroscience, UCL. Anchors central nervous system and blood brain barrier applications, where single domain formats are especially valuable.

Advisory bench

Senior antibody engineering expertise is a known gap. We close it through scientific advisors and a senior antibody scientist hire, treated as a first-class priority rather than an afterthought. Commercial sits as its own named seat rather than being absorbed by the scientific ones.

Nanobody engineering and developabilityScience · Recruiting

Antibody informatics and profiling methodsScience · Recruiting

Structural bioinformatics and analysis toolingScience · Recruiting

Commercial, pricing and go to marketCommercial · Recruiting

Contact

Tell us the target and the idea.

A screening report answers the question most customers have first, and it is approved without a board conversation. The campaign follows the answer. We are also talking to senior antibody scientists who want to run a wet lab programme from the first campaign onwards.

Abhi · Technical abhi@lambda-tech.org

Alexander · Scientific alexander@lambda-tech.org

References

Chai Discovery, Chai-2, bioRxiv, July 2025.
Germinal, Arc Institute, bioRxiv September 2025 and Nature Biotechnology 2026.
Swanson et al., mBER, Manifold Bio, bioRxiv, September 2025.
The Antibody Society, Antibodies to Watch, 2025.
Mordor Intelligence, antibody discovery outsourcing, 2025.
ProteoGenix campaign cost guidance and practitioner-reported phage pricing, converted to sterling.