# Drugverse Drugverse joins public biomedical databases into one versioned and queryable data layer. Then it builds knowledge graphs and entity views on top of that layer. ## Problem Scientists find it hard to parse and aggregate biomedical databases. - Each database has a different format. Examples: DrugBank, Open Targets, HGNC. There are many more. - The databases do not talk to each other. So enrichment and cross-referencing across sources is lost. - Current options are bad: - **License a parser from one vendor.** It is expensive. It covers only that vendor's data. - **Build your own in R.** It is complex and slow. ## Solution Drugverse has two products: **DB Parser** and **Scope**. ### DB Parser DB Parser converts many databases into one universal data object. - The object is: - **Queryable.** - **Versionable.** New source versions flow through. - **Scientifically integrated.** It uses the cross-reference IDs in each source. Example: DrugBank says an entity also has ID *X* in another database. DB Parser uses these links to join entities into one object. - Output is an R dataset. Size: ~50–500 MB for a typical set of sources. ~1.7 GB with all sources. - Drugverse loads the R dataset into **Postgres**. Postgres is the single source of truth. - A **knowledge graph** sits on top of Postgres. See [[knowledge graphs]] and [[ontologies]]. - It focuses on one therapeutic area, gene, disease or target. - It pulls new evidence from unstructured sources (PubMed, NCBI, the web). It analyses that evidence and links it to existing data. - Update loop: new source versions → parse → enrich → update Postgres → update the knowledge graph. The graph stays current. - **DB Parser Pro** joins a client's own databases with the public databases. It runs on the client's server. No data leaves the client. ### Scope Scope gives full information on one biological entity: a gene, disease, target or drug. - It resolves ambiguity in the data. Examples: a gene symbol that is also a protein symbol, an old symbol, a renamed drug. - For each entity, it shows: - Source of each fact (example: Open Targets) and its type. - IDs of the entity in other databases. - Other names. - Mechanism. Example: which targets aspirin affects. The user can filter the view to one system, such as the immune system. - Scope's graph view is a primer. The knowledge graph goes much deeper. - Main users: bioinformaticians. General users can also use it. ## Differentiation - **Multi-source.** Vendor parsers cover one source. Drugverse joins all of them. - **ID-level integration.** Entities join through the sources' own cross-reference IDs. It is not a loose text match. - **Versioned.** Each source version is tracked and the update flows to the graph. - **On-premise option.** DB Parser Pro keeps private data on the client's server. - **Agent-ready.** Postgres plus a knowledge graph is a clean target for AI agents. See [[Context Layers MOC]]. ## Value - Less time and cost to build an integrated dataset. No custom R pipeline. No set of single-vendor licences. - More insight per query, because cross-source links are now visible. - A private-plus-public view of the data, without data exposure. - A knowledge graph that stays current for one disease, gene or target. ## Opportunities - **Drug repurposing.** Scope maps drugs to targets and mechanisms across sources. See [[Drug Discovery]]. - **Agent querying.** Agents query the Postgres layer and the knowledge graph directly. - **Niche intelligence products.** One continuously updated graph per therapeutic area, gene or disease. - **Enterprise integration.** DB Parser Pro for pharma and biotech teams with private data. - **Research to product.** Drugverse built one tool with an Australian research team. The team agreed to turn its research into a product. > [!question] Open points > - Full list of supported databases. The transcript names DrugBank, Open Targets and HGNC. Other names were unclear. > - Scope build status. One tool exists with the Australian team. The product is not built yet. ## Go-to-market notes - [[Drugverse RL Environments and Data Play]] - [[Biotech Neo Labs - Buyer List]] - [[Building ChatGPT Apps and Claude Connectors]] - [[Making Drugverse Agent-Accessible]] ---- tags: #boat #deeptech links: [[TechBio MoC]] [[Drug Discovery]] [[knowledge graphs]] [[ontologies]] [[Context Layers MOC]] [[Knowledge Graphs for Industrial Data]] [[Identified Opportunities in Medicine Development]] [[Digitalisation of Biology]]