Drug Discovery AI
D

DrugBank

DrugBank sells Biopharma OS, an intelligence layer for pharmaceutical research, portfolio strategy, business development and competitive intelligence teams, built on the curated drug knowledgebase that began as a University of Alberta research project 20 years ago. The knowledge graph connects more than 150 million data points on drugs, targets, mechanisms, diseases, trials and companies, including more than 230,000 trials and 9,000 company pipelines. It resolves sponsors to parent organizations, maps identifiers across more than 20 ontologies and traces each record to its source.

Teams work in three ways. The Workspace lets them save a target or asset landscape, annotate results and set alerts, such as new KRAS trials in the past two weeks. A Mapping Agent resolves names to graph entities, a Monitoring Agent tracks changes and a Validation Agent checks source, mechanism and trial data before an answer goes out. The API, GraphQL and the DrugBank MCP server put the graph inside Claude, ChatGPT or a custom agent, and a Microsoft 365 Copilot plugin brings it into Copilot Chat, Word, Excel and Teams. Enterprise customers can bring proprietary data and run the graph and agents in their own environment. A separate Clinical Intelligence API serves drug information and interaction checks to EHR and telehealth developers.

DrugBank data is used at more than 1,500 research institutions and cited in more than 65,000 publications, with Servier, Bayer, Healx and the NIH named among its users. Lisa Downey has been chief executive of the Edmonton company since October 2025, with cofounder Mike Wilson as chief product officer. Commercial plans are priced on usage, and verified academics get a limited free research edition.

AI Health Index verifiedNovember 19, 2026
Compare DrugBank with other vendors
Founded
2015
Headquarters
Edmonton, Alberta, Canada
Website
go.drugbank.com
Categories
drug-discovery, clinical-trials-ai
Indexed Products
Biopharma OS, DrugBank Knowledge Graph, Workspace, DrugBank MCP, Mapping Agent, Monitoring Agent, Validation Agent, Clinical Intelligence API
Buyer Segments
Biopharma R&D teams, Portfolio strategy and business development teams, Academic researchers, Health technology developers
Assessment

Capability Axes

An AI Health Index grade measures what a buyer can verify from public sources on the date shown. It is not a rating of how good the product is. A vendor can build an excellent system and grade low on an axis because it publishes nothing an outsider can check. How grades read

AI Capability
CC on AI CentralityArtificial intelligence is a feature layer on a product whose value stands without it. A learned component is a model whose parameters are estimated from data: machine learning, or a statistical model fitted to population data, including one then updated on the patient measurements. A fixed formula, a rule set, or a feedback controller that adjusts a dosing factor from the patient own readings is not one, however patient specific its output and whatever the vendor calls the adjustment.
Vendor Published

Agents sit on top of a curated graph. The Mapping Agent resolves a name or identifier to an entity, the Monitoring Agent follows changes to drugs and trials, and the Validation Agent checks source and mechanism information before delivery, while customers can also drive the graph from their own model client over the API or MCP.

The graph does much of the work without them. Curated records and directional relationships across pharmacology, trial and company data answer structured queries directly, and deterministic queries and aggregation tools produce the trial counts in the company's own demonstration. Which steps in the named agents use a learned model, as opposed to graph queries, curated mappings and fixed rules, is not explained.

DD on Autonomy and Oversight ModelNo oversight structure is published. An absolute claim that the system does not err grades here too, because a buyer who believes it will not build the review step that would catch a failure.
Vendor Published

Researchers stay in charge of what they act on. They set alert thresholds, save investigations, open the sources behind any answer, annotate results and share projects with colleagues, while the Validation Agent checks source, mechanism and trial data before delivery and curators review what enters the graph.

The Biopharma OS page promises a platform built so researchers never miss the signals that matter. Review can catch a wrong answer but not a trial the agent failed to surface, and no step puts a person between the agent's search and its answer to check for omissions. There is no described threshold for sending an unresolved result to a person.

BB on Model and Technology TransparencyThe approach or the suppliers are named without the version and update discipline behind them. Naming a supplier is the entry to this band both here and on Model Supply Chain Disclosure, which ask different questions of the same disclosure: who receives the data, and what produces the output.
Vendor Published

The intelligence layer combines the curated graph with entity mapping, deterministic graph queries and aggregation tools, and a harmonization process resolves sponsor names and links records across more than 20 ontologies. Answers carry the data release they came from. Researchers can query through MCP from Claude, Copilot, ChatGPT or a custom agent, and the Workspace offers model assisted or manual analysis. The MCP layer, built under the European ITEA 4 AIDESL project, works with any compatible client.

The model family and version behind the built in Mapping, Monitoring and Validation agents is not named, nor is how their training data was chosen or how a model change reaches a deployed Workspace.

CC on Model Supply Chain DisclosureThe architecture is described and no model provider is named. Naming a hosting provider alone does not lift a record out of this band. Record the host in the note, because it matters for residency and breach scope, and grade on the model layer, which is the question this axis is named for.
Vendor Published

Google hosts the service, Amazon Web Services handles storage and backups and Cloudflare delivers traffic, according to the subprocessor schedule, which also names Anthropic for assistant content submitted by DrugBank's own employees. Customers who bring their own client, whether Claude, Copilot or ChatGPT over MCP, keep the downstream model under their own arrangement.

The model provider behind the built in Mapping, Validation and Monitoring agents is not identified, and neither are the training and retention terms on that path. Enterprise customers who host the agents internally control that layer themselves, while in the hosted Workspace the question stays open.

CC on Clinical and Operational EvidenceNamed customers, or vendor reported percentages with no method, denominator or reference standard. Scale of use is recorded here and is not treated as evidence of benefit.
Vendor Published

The company's own test, run in September 2026, asked the same trial landscape question three times of an agent with and without DrugBank connected. With the graph, each run returned 131 trials, 29 sponsors and four classified stop reasons. Without it, the runs returned 40, 31 and 12 trials. That shows repeatability, but no independent reference set says which trials a complete answer should include.

Use of the underlying data is broad, with more than 1,500 research institutions and more than 65,000 citing publications, and Healx and the Milner Therapeutics Institute among named research users. None of that evaluates the agents on the research tasks they are sold for.

CC on AI Safety and PHI StewardshipGeneral assurances of privacy and security that do not answer the questions artificial intelligence raises: what is retained, what reaches a model, and what happens to it there.
Vendor Published

Enterprise customers can connect proprietary data to the graph. Data is stored encrypted on Google Cloud in the United States, with a customer hosted option for organizations that want it inside their own environment. The service privacy policy keeps information as long as the service and certain legal or business purposes need, and rights to train models on the licensed graph are an enterprise contract matter.

Whether customer uploads or prompts train the hosted agents, how long prompts and workspaces are kept, and how identifiable study data would be handled are not committed to in the standard product. Most of what flows through the platform is drug, trial and company information rather than patient records.

Regulatory and Compliance
CC on HIPAA and BAA PostureA status is stated but not supported: compliance is claimed, or business associate status stated, without the underlying document, including where the only privacy notice published covers the website rather than the service that handles patients. The mirror case grades here as well: a substantive privacy document that reaches the service, with no statement of business associate status anywhere and no scope position taken. One of the two elements the band above asks for is present, which is a grade below it and not a grade at the floor.
Vendor Published

The research workspace runs on compound, trial and company information, though organizations can connect their own data, and the separate Clinical Intelligence API serves medication reference and interaction workflows inside clinical software. The privacy policy reaches information entered through the services and places storage and processing in the United States.

The underlying data centers are described as meeting HIPAA requirements, while DrugBank lists HIPAA among certifications it is still evaluating for itself. It makes no statement that it acts as a business associate for either product and offers no agreement tier, which matters most for developers who send medication lists from patient charts to the clinical API.

DD on Security Certifications and Trust CenterControls are asserted with nothing independent behind them, or nothing is published. Read the note before concluding anything: this is the grade most often corrected on a second pass, because assurance material frequently sits on a parent domain or inside an old announcement rather than on the product pages.
Vendor Published

Engineering controls include mandatory peer review for code changes, automated testing before deployment, staged rollout and rollback, encrypted network traffic, isolated databases and customer data encrypted at rest. Hosting on Google Cloud brings that provider's certifications to the infrastructure layer, and customers under agreement can see further operating evidence, including records of recovery exercises.

DrugBank says its own practices align with SOC 2 controls and that certification is not yet complete. No independent assessment of its own operations and no external penetration testing cadence has been reported, so the assurance a buyer can verify today stops at the cloud provider.

CC on FDA and Regulatory StatusNo device claim is made and the product is scoped accordingly. Most administrative and operational products sit here and are not penalized for it, because this axis grades the appropriateness of the positioning rather than possession of a clearance. A stated limit on use, such as research use only or a clause that the product gives no medical advice, beside marketing that sells the product for clinical decisions also sits here, with both quoted, because the buyer has to ask which one governs.
Vendor Published

Biopharma OS is sold to research, portfolio and business development teams to evaluate targets, assets, trials and portfolios, upstream of any clinical use. The graph carries FDA, EMA and Health Canada labels as sources of drug information, and those approvals describe the medicines, not the software.

The Clinical Intelligence API sits closer to care. Customers embed its medication information and interaction checks in clinical software, where the output can reach a prescriber. Neither product states a device position or names the exemption it relies on, and the clinical API's interaction checks carry no stated intended use.

CC on AI Governance and Bias DisclosureResponsible artificial intelligence is committed to in policy language with no evaluation behind it. Most of the index sits here.
Vendor Published

Quality control is built around the data. Curators and biostatisticians review the graph, the Validation Agent checks sources and trial data before answers go out, and every record traces back to its source with the mechanism or association shown and trial registry identifiers attached.

DrugBank describes its connected answers as unbiased and deterministic, and its demonstration repeats one landscape question three times with identical results. That shows repeatability on one target, not coverage. How well answers hold across therapeutic areas, countries and sponsor types, and whether the graph underrepresents some sources, has not been measured publicly. An advisory board formed in May 2026 guides AI adoption, with no stated role in model governance.

CC on AI Liability and RecourseMechanisms exist that let someone challenge an output, such as audit trails, source traceability or review before commit, with nothing standing behind the output and no route for the harmed party.
Vendor Published

A research answer can be checked line by line. Results link to original sources with identifiers for trials, drugs and sponsors, saved Workspace projects can be reopened and annotated, and alerts carry a source that colleagues can verify, so a team can challenge an answer and reconstruct the basis for a decision. Each answer also carries the data release it came from, such as release 1423972 in the GLP-1 half life example, which pins a result to a dated version of the graph.

The standard commercial terms offer no correction process for a wrong research conclusion and no remedy for a portfolio or licensing decision built on an omitted trial. The privacy terms provide correction rights only for personal data.

Integration and Deployment
CC on EHR and Interoperability DepthIntegration is claimed through standards or a middleware layer with no system named and nothing to verify.
Vendor Published

Biopharma OS connects through an API, GraphQL and an MCP server, and the Microsoft 365 Copilot plugin carries it into Word, Excel, Teams and the Researcher agent. Identifier mapping across ChEMBL, PubChem, RxNorm, ATC and UNII lets a research team carry a reference from its own data into the graph.

Clinical record systems are reached only through the separate Clinical Intelligence API, which returns drug concepts and interaction data as structured JSON for EHR and telehealth developers to embed. No named EHR connector, and no production deployment reading from or writing to a record system, is described for either product.

BB on Deployment Model and Data ResidencyOptions and residency are stated with isolation or the processing path left open.
Vendor Published

Academic and Team editions run in DrugBank's hosted environment. Enterprise customers can choose cloud, hybrid or internally hosted deployment, with the graph and agents running inside their own environment. The privacy policy places servers in the United States, and the subprocessor schedule names Google for hosting, Amazon Web Services for storage and backups, and Cloudflare for traffic handled at the nearest global location.

For the standard hosted tier, how tenants are isolated and what residency commitment covers proprietary data and agent requests are not stated, which leaves the internally hosted route as the one with a location the customer defines.

Commercial
BB on Commercial TransparencyA price or a pricing basis is published without full tiers, so a buyer can size the cost before making contact.
Vendor Published

Team and Enterprise plans meter intelligence work across the Workspace, API and MCP rather than charging by seat. Team includes the graph, the agent toolkit and a focused set of active analyses. Enterprise adds capacity, deeper proprietary data integration, cloud, hybrid or internally hosted deployment and model training rights by agreement. Existing Biopharma OS customers on Microsoft 365 Copilot can add the Copilot plugin with no extra purchase, and verified academics get limited tools and downloads under a noncommercial license.

There is no entry price, no rate for converting analyses into billed usage and no stated minimum commitment.

CC on Setting and Specialty CoverageCoverage is claimed broadly without specifics, or stated clearly with nothing validating it yet.
Vendor Published

Pharmaceutical research, portfolio planning, licensing and competitive intelligence are the settings. Teams explore target biology, compare assets and trial programs, and watch for sponsor announcements across more than 230,000 trials and 9,000 company pipelines, and data science groups use the resolved data to train their own models. Healx and the Milner Therapeutics Institute have used it for rare disease and wider therapeutic research.

The September 2026 agent test covers one target and one landscape task. How the agents perform across disease areas, development phases and organization sizes has not been shown, and the clinical API serves a different buyer, the software developer, from the rest of the product.

Alternatives

Alternatives to DrugBank

Other products that do the same kind of work as DrugBank, and what each one does.

Causaly supplies scientific AI software for pharmaceutical, biotechnology and academic research. Its agents combine biomedical literature, knowledge graphs and an organization's internal evidence to investigate questions, compare findings and produce assessments.

BenchSci develops EMET, an AI research environment for scientists in biopharma and academia. The platform combines scientific literature, experimental databases and internal research with a biomedical knowledge graph.

Medvolt, based in Pune, India, sells MedGraph, an AI and physics based in silico platform for early drug discovery, to pharmaceutical and biotechnology companies as software modules and as delivered projects.

Edison Scientific is the commercial spinout of the nonprofit research lab FutureHouse, formed in November 2025 to run its AI scientist platform.

LatchBio runs a cloud workspace where biotech, pharma and academic research teams store, analyze and visualize omics data, alongside an AI research lab that benchmarks how well agents analyze biological data.

Immunai occupies a different position from most of this category. It is not developing its own drugs.

Commercial

Pricing

Vendor-published figures are labeled as such. Figures labeled “Estimated” are derived from third-party sources and have not been confirmed by the vendor.

Entry Price Pricing Basis BAA Tier Implementation Source
Usage based custom quote
Verified academics can access limited online tools and downloads for noncommercial research. Team and Enterprise meter intelligence work across the workspace, API and Model Context Protocol rather than charging by seat. Enterprise adds deeper integration and more deployment options. The service privacy policy governs information submitted through DrugBank services. No business associate agreement is stated for the research platform or for the separate clinical API. Enterprise supports proprietary data harmonization and cloud, hybrid or customer hosted deployments. Implementation scope varies with the systems involved, and no setup or integration fee is listed. Vendor Published

Team and Enterprise include commercial use, and Enterprise model training rights are available by agreement. Existing customers on Microsoft 365 Copilot can add the Copilot plugin with no extra purchase. Commercial entry prices, the conversion of activity into billed usage and minimum commitments are set in the quote.