Structure-informed diligence and target intelligence
Offer 1 module under Portfolio Intelligence System: one independent diligence package that answers: Is the biology credible on public data? What do approved drugs in this class actually look like? And if you are showing a protein or complex, where is that story strong vs speculative? We write the cross-layer synthesis—not three separate PDFs—so startups and investors can move IC, BD, and pipeline conversations with explicit limits on what the data can support.
For strategic and due-diligence use only—not regulatory, medical, or investment advice. Structure outputs are computational models, not experiments. Open Targets and Orange Book are snapshots of public data at the time of the pull.
What problem this solves
| Situation | What goes wrong | What we deliver |
|---|---|---|
| Deck says “validated structure” or “clear complex” | A fold or cartoon is treated like crystallography | A confidence-aware read: where the model is reliable, where it is not, and what to validate before partners or IC rely on it |
| Target story is heavy on biology, light on commercial reality | Associations without “what is already approved?” | Orange Book–anchored precedent: what listed products exist in the class (comparators—not a prediction your asset will be approved) |
| Diligence is a pile of screenshots | Open Targets, OB, and structure treated as unrelated | One executive narrative that states alignments, tensions, and open questions across layers |
| You need an outsider who does not invest | Conflicted or opaque sourcing | Traceable public inputs (queries, listing scope, software lineage) in a methods appendix |
Outcomes you can use this week
- Investment committee — A short downside / overclaim section you can drop into a memo: “What we would ask the company” and “What public data does not show.”
- Founder / BD — Talking points that separate evidence (OT), approved precedent (OB), and structural hypotheses (AlphaFold-class) so you do not over-promise in partner meetings.
- Pipeline choice — A single ranked view of “what public data supports now” vs “what requires new data” before the next capital or partnership gate.
How an engagement runs
- Scoping call — Entities (target / disease / drug), whether you need Orange Book and/or structure-informed work, and sequences if folding is in scope.
- Public pulls + optional structure run — Open Targets evidence via the same reproducible approach as How it’s built; FDA Orange Book scoped to your comparator question; optional open-source AlphaFold 2 pipeline for structural hypotheses (code Apache-2.0, parameters CC BY-4.0—attributed in the appendix).
- Written package — Executive brief, cross-layer synthesis, risk/overclaim section, and methods appendix. Ranked models or listing notes are supporting artifacts, not the product.
Illustrative (IaIP blood product): For a proposed IaIP protein therapeutic, Open Targets grounds ITIH-linked indication and mechanism evidence and adjacent drug/candidate context; Orange Book supplies listed approved-drug precedent where it overlaps the question, with an explicit call-out when plasma-derived or BLA-class products are thin or absent in OB so diligence does not misread the market. Structure-informed reads apply if product or complex sequences are in scope—see the same thread across offers.
The three public layers (and what each is for)
In a fused engagement we integrate these—not stack them without interpretation.
Open Targets — “Is the target–disease–drug story supported?”
The Open Targets Platform is the integrated evidence graph (associations, genetics, tractability, safety signals, expression, and more)—same resource described in The platform.
You get answers to: How strong is public support for this target in your indication? What does the competitive drug–target picture look like? What do tractability and safety signals suggest for modality?
Orange Book — “What does approved reality look like in this class?”
The Orange Book lists FDA-approved products (and related listing context). We use it for comparator and differentiation narratives—what is already on the market—not to forecast approval of a new asset.
You get answers to: Which approved products matter as precedents for your story? What does “crowded” vs “differentiated” mean against listed drugs?
Open-source structure prediction — “What can we responsibly say about shape and assemblies?”
We use the reference AlphaFold 2 inference pipeline so structural reads are reproducible and attributable. Outputs include domains, interfaces, multimer hypotheses, and plain-language confidence (e.g. pLDDT / PAE)—not a substitute for OT or OB facts.
You get answers to: Where is the model trustworthy vs weak? Do slides outrun the evidence? What experiments would a sceptical IC or partner ask for next?
AlphaFold is for theoretical modeling only and is not for clinical use; see the repository disclaimer.
What you receive (deliverables)
- Executive brief (typically 5–10 pages) — Decision-oriented; integrates OT, OB (if scoped), and structure-informed reads (if scoped).
- Competitor analysis (structured) — Open Targets: disease- and target-level drug/candidate density, mechanism class, and pathway neighbours. Orange Book: listed approved comparators and RLD context where relevant. Structure-informed (if scoped): how modelled protein geometry supports or weakens differentiation vs other biologics or extracellular targets (confidence explicit).
- Investment / IC diligence framework — Not buy/sell advice: a memo-ready layout—qualitative go / no-go / dig deeper, thesis vs risks, competitive threats, catalysts, and suggested follow-up diligence—each tied to which public layer supports the line (OT vs OB vs structure).
- Explicit decision enablement — Where useful, decision gates (if/then), owner–action tables, and dataroom checklists so IC and management can record what public data does and does not decide (section structure varies by scope).
- Risk / overclaim appendix — Bullets you can reuse in IC or dataroom Q&A.
- Optional slide-ready figures with uncertainty called out.
- Methods appendix — Open Targets query scope; Orange Book search scope and date; structure software version and presets (monomer vs multimer, database preset).
What this is not
- Not experimental validation, CMC advice, or regulatory strategy.
- Not a guarantee that a structure or target will succeed clinically or commercially.
- Not “exclusive access” to AlphaFold—we sell interpretation and a documented workflow on top of public tools and data.
Who it is for
Startups — Tighter fundraising and BD narratives: what to claim now vs what to prove next, with evidence + precedent + structure aligned.
Investors — Faster go / no-go / dig deeper with a single writer synthesising OT + OB (+ structure) instead of reconciling three streams yourself.
See also For startups and For investors.
Trust and transparency
- Open Targets — How it’s built describes the GraphQL, reproducible workflow.
- Orange Book — Public FDA data; pull logic documented in your appendix.
- AlphaFold 2 — google-deepmind/alphafold; cite per upstream guidance.
Illustrative or mock examples in any materials are labeled as such.
Next step
Contact with: Open Targets–only, Open Targets + Orange Book, or Open Targets + Orange Book + structure-informed (include entities, sequences/FASTA and stoichiometry if relevant, and timeline). We reply with a fixed fee and delivery date when scope is clear.