The Biotech Intelligence Stack
Seven workflows where genomics, drug discovery, pipelines, patents, deals, and revenue run on one source-attributed knowledge graph, queryable from any AI client. The consolidation of the tools you currently buy separately, at a fraction of the price.
Nexotype Team · June 22, 2026
Key takeaways
- 1. One graph, not five subscriptions. Genomics, drug discovery, pipelines, patents, deals, and revenue live on a single biomedical knowledge graph. The traversal from a variant to a revenue line is one query.
- 2. Every claim has a source. A source and an evidence record sit behind each fact, tied to PubMed, ClinicalTrials.gov, DrugBank, or a patent filing. If it is not in the graph with a source, the engine cannot invent it.
- 3. Commercial intelligence is a first-class layer. Seventeen tables cover companies, patents, M&A, licensing, pipelines, the patent cliff, revenue, market sizing, and catalysts, not a bolt-on to a biology database.
- 4. Queryable from any AI client. A remote MCP server exposes seven query tools to Claude and other agents, with no install. No incumbent ships one.
The landscape, and the gap
Biotech intelligence is split across point tools, each strong on one axis and silent on the rest. They are also priced for the enterprise. The wedge is the connected graph plus grounded AI, at a price the underserved buyer can actually pay.
| Tool | Strong on | Where it breaks |
|---|---|---|
| Citeline (Norstella) | Pipeline + trial + competitive intel, deep coverage | $50k to $200k a year; report-oriented; no genomics layer; no AI query |
| Cortellis (Clarivate) | Pipeline, IP, regulatory, and safety modules | $20k to $150k a year; modules priced and queried separately; tabular, not a graph |
| Evaluate Pharma (Norstella) | Consensus drug sales forecasts and market intel | $30k to $100k a year; forecast-first; no variant or target biology |
| GlobalData Healthcare | Pipeline, epidemiology, and sales data | $20k to $80k a year; weak cross-entity traversal; no grounded AI layer |
| DrugBank / Open Targets | Drug, target, and gene biology, often free | No commercial layer at all: no patents, deals, revenue, or market sizing |
| Nexotype | One source-attributed graph from genomics to revenue, queryable via MCP | Coverage is growing toward the incumbents and expands every month, on one connected graph with grounded AI at a fraction of the enterprise price |
Why a biotech team is the one writing this
Every pick here ties back to a mechanism: cortisol, SHBG, AMPK, mitochondria, glutathione. That is the exact territory we map at Nexotype, a biomedical intelligence platform that connects compounds to the genes, proteins, and pathways they act on. This guide is the consumer edge of the same knowledge graph.
Open the platformThree threads, seven workflows
Discovery and graph
The biology surface: traverse from a variant to a gene to a drug to a trial, find every program against a target, and surface repurposing paths through shared pathways.
Commercial intelligence
The money surface: the patent cliff math, the full company deep dive, and the deal and licensing landscape, all as structured rows rather than PDFs.
AI-native
The query surface: the same graph, asked in plain English from Claude or any MCP client, with every answer traced back to its source.
What Nexotype delivers
One knowledge graph. Sixty-one entity types across ten domains, from genes and proteins to drugs, trials, patents, companies, and revenue lines, connected by typed relationships rather than stored in isolated tables.
Grounded answers. A source and an evidence record are the spine. Every node and edge can name where it came from, so an answer is auditable and the AI layer cannot fabricate a fact the graph does not hold.
Commercial intelligence. Companies, patents, deals, licensing, pipelines, regulatory approvals with the patent-cliff exclusivity date, revenue lines, market sizing, and the catalyst calendar, all on the same graph as the biology.
Variant to drug in one traversal
A single path runs from the mutation a patient carries to the trial that might treat it.
DrugBank and Open Targets describe the biology beautifully and stop there. Cortellis starts at the commercial end and rarely reaches the variant. Nexotype keeps the genomic and the commercial on the same graph, so the traversal from a gene to the drug that targets it to the trial it is in is one query, not a manual join across two subscriptions.
Nexotype
- One traversal: gene to protein to mechanism to drug to trial
- The drug-target link joins the biology to the commercial catalog
- Every edge carries a source and an evidence record
- Same graph holds the patent and the company on the other side
Biology-only tools
- Excellent target biology, but the trail ends at the drug
- No patent, deal, revenue, or market layer to continue into
- Commercial context requires a second, separate subscription
- Joining biology to commerce is a manual export-and-merge
One graph that holds both the variant and the revenue line is the difference between an answer and a research project.
Every program against a target
Who is developing a GLP-1 drug, and how far along is each one.
Asking who targets GLP-1 across an incumbent platform means opening the pipeline module, the company module, and the deals module and reconciling three exports. Nexotype answers it as a single competitive-landscape traversal: from the target to every mechanism that hits it to every program and the company behind each, ranked by clinical phase. Novo Nordisk and Eli Lilly anchor the list, with a long tail of fast followers.
Nexotype
- One competitive-landscape call from target to companies
- Ranked by clinical phase from the pipeline records
- Crosses biology to commercial without leaving the graph
- Available on the free tier and to any MCP client
Enterprise suites
- Pipeline, company, and deals priced as separate modules
- A landscape view means reconciling three exports by hand
- Twenty to two hundred thousand dollars a year to start
- No agent can ask the question for you
A landscape is a graph question. Tools that store it in separate tables make you do the join they should have done.
Drug repurposing through shared biology
Start from a drug, walk to its pathway, and find the other diseases that pathway touches.
Report-oriented platforms answer repurposing with a literature search and a human to read it. Nexotype answers it structurally: a drug links to a target, the target sits on a pathway, and the pathway connects to every other indication with shared biology. Semaglutide, anchored at the GLP-1 receptor and the insulin signaling pathway, walks naturally out from diabetes toward obesity, cardiovascular risk, and metabolic liver disease, exactly the expansion the real program took.
Nexotype
- Pathway membership makes repurposing a graph walk
- KEGG and Reactome identifiers anchor each pathway
- Ranks candidate indications by shared biology, not keyword overlap
- Reproducible: the same traversal gives the same candidates
Report-first tools
- Repurposing reduces to a literature keyword search
- A human reads abstracts to infer the connection
- No structured pathway layer to traverse
- Hard to reproduce or audit the reasoning
Repurposing is a shape in the graph. You either store the shape or you reconstruct it by hand every time.
The patent cliff, as one date
When does the exclusivity actually end, and how much revenue is standing on it.
Every analyst has rebuilt a patent cliff in a spreadsheet, chasing the base patent, the term extension, the pediatric add-on, and now the Medicare negotiation list across four sources. The effective exclusivity end is that math, stored. For a drug like Humira it rolls the whole stack into the single date the revenue actually falls off, and the revenue lines on the other side say how much is standing on it. The biosimilar count says how fast it erodes.
Nexotype
- The effective exclusivity end is the bottom-line loss-of-exclusivity date
- Folds patent term, extensions, pediatric, orphan, and IRA into one date
- Whether biosimilars have launched, and how many, signals the erosion speed
- The revenue lines tie the date to the dollars at risk
Spreadsheet workflow
- Base patent, PTE, and pediatric chased across separate sources
- IRA negotiation status tracked in yet another place
- Revenue pulled from a 10-K PDF and keyed in by hand
- Rebuilt from scratch for every drug, every quarter
The cliff is a calculation everyone redoes. Storing it as a field turns a recurring spreadsheet into a lookup.
A whole company in one call
Portfolio, pipeline, patents, deals, and revenue for a company, without opening five modules.
On an incumbent suite a company profile is assembled by hand from the pipeline module, the IP module, the deals module, and a financial export. The company record holds the spine, and a company deep dive returns the rest in one response: the owned assets, the development pipeline, the patents assigned, the deals and licensing agreements, the reconciled revenue lines, and the technology platforms. Subsidiaries link through a parent-company field, so the corporate tree comes with it.
Nexotype
- A company deep dive returns the whole entity in one response
- Portfolio, pipeline, patents, deals, revenue, and platforms together
- A parent-company field links the subsidiary tree
- Same data the MCP layer serves to an AI client
Module-priced suites
- Pipeline, IP, deals, and financials are separate products
- A company profile is a manual assembly across four exports
- Each module carries its own license and its own query language
- The corporate hierarchy lives in a different place again
A company is one node with everything hanging off it. Pricing each branch separately is a packaging choice, not a data one.
The deal and licensing landscape
Who is buying and licensing what, with the structure of each deal as data.
A deal in most databases is a headline and a total. A deal record stores the structure: upfront cash, maximum milestones, royalty percentage, and equity stake, with a rationale field for the strategic logic. A licensing agreement carries the same shape plus territory and status. Around them, the catalyst calendar holds the JPM and ASCO moments where deal flow clusters, linked to the companies presenting.
Nexotype
- Deal structure as fields: upfront, milestone, royalty, equity
- Licensing carries territory, status, and term
- A linked catalyst calendar of conferences and readouts
- All queryable next to the companies and assets involved
Headline databases
- A deal is a total value and a press-release summary
- Structure lives in a PDF, if it is captured at all
- Catalysts tracked in a separate calendar product
- No structured link from the deal to the asset or the company graph
Deal structure is the analysis. A platform that stores only the headline has thrown away the part you needed.
Ask the graph from Claude
Show me Novo Nordisk diabetes drugs and the genes they target, answered in chat with citations.
No incumbent ships an MCP server. Cortellis has an API; nobody else lets an agent ask the question. Connect Claude Desktop to mcp.nexotype.com, type the question in plain English, and the deterministic query engine answers from the graph: Semaglutide and Liraglutide for type 2 diabetes, both targeting the GLP-1 receptor on chromosome 6, in the insulin signaling pathway, with the PubMed, ClinicalTrials.gov, and DrugBank identifiers attached. If a fact is not in the graph with a source, the engine cannot invent it.
Nexotype
- Remote MCP server, zero install, paste one URL
- Seven query tools over the same graph the app uses
- Every answer carries its PubMed, NCT, or DrugBank source
- A deterministic engine, not a chatbot guessing over PDFs
Everyone else
- No MCP server; at best a REST API a developer must wire up
- No agent-native access to the data at all
- AI features, where they exist, summarize documents and can drift
- No source attached to the generated answer
Agent-native, source-attributed access to biomedical and commercial data is a category nobody else is in yet.
What links these seven
None of the workflows above is a separate product. They are seven views of one graph.
- One graph. The variant, the drug, the trial, the patent, the deal, and the revenue line are nodes on the same graph. The work is the traversal, not the integration.
- One source-attribution model. A source and an evidence record ground every node, so the same audit trail runs from the genomics layer to the revenue line.
- One query layer. The app and the MCP server read the same seven tools, so what you can click you can also ask an agent to do.
Run a traversal yourself
Open the platform and run a company deep dive, or connect any MCP client to the server and ask in plain English. The free tier covers it.
https://mcp.nexotype.com/mcpHow it works under the hood
The graph spans sixty-one entity types across ten domains: omics, clinical, assets, knowledge graph, LIMS, engineering, commercial, financial, standardization, and the personal-health layer. Every record carries the same two foundations: an audit base with timestamps, soft delete, and created, updated, and deleted authorship, and an ownership marker that separates the curated shared graph from per-user private data. Nothing is a black box; every row knows when it changed and who owns it.
The AI layer is a deterministic query engine over that graph, not a language model improvising over documents. A question resolves to a traversal across the seven tools, and the answer is assembled from the nodes it visits, each able to name its source. That is what grounded means here: the engine returns what the graph holds, with the citation attached, and declines what it does not.
What could go wrong
- Curated coverage is smaller than the incumbents. Citeline and Cortellis have decades of breadth. Nexotype trades breadth for one connected graph, grounded answers, and a price a small team can pay. Coverage is expanding; it is not yet exhaustive.
- The genomics layer is research and educational. Variant and biomarker analysis is for exploration and intelligence, not clinical diagnosis. It informs a thesis, it does not replace a clinician or a validated diagnostic.
- Some surfaces are still maturing. The MCP layer and a few commercial fields are early. Where a number is a projection or a snapshot rather than a reconciled filing, the data says so.
Why a biotech team is the one writing this
Every pick here ties back to a mechanism: cortisol, SHBG, AMPK, mitochondria, glutathione. That is the exact territory we map at Nexotype, a biomedical intelligence platform that connects compounds to the genes, proteins, and pathways they act on. This guide is the consumer edge of the same knowledge graph.
Explore the platformWritten by the Nexotype Team. Model names and fields refer to the live data model; worked examples use publicly known facts and identifiers verified against public sources (PubMed, ClinicalTrials.gov, DrugBank, FDA filings).
Nothing here is investment, medical, or regulatory advice. Commercial figures are summaries of public data and, where marked, computed projections rather than reported filings. The genomics layer is for research and education, not clinical diagnosis.