Company Dossier · Cambridge, MA
Plex Research
The Cambridge company that turned a search bar into a drug discovery engine - and made the AI show its work.
The Cambridge company that turned a search bar into a drug discovery engine - and made the AI show its work.
Plex Research asks a deceptively simple question of the drug discovery world: what if the search bar scientists use every day could reach past the summaries and into the raw experiments themselves?
Founded in 2017 and based at 625 Massachusetts Avenue in Cambridge, Plex Research builds an AI-powered scientific intelligence platform for drug discovery. At its center is a biomedical knowledge graph that integrates billions of minimally processed experimental data points - genomics, proteomics, compound activity, clinical data - into a single structure a scientist can query in plain language.
The company is explicit about what it is not. "Plex is not a literature search tool," its own materials note. Rather than summarizing published papers, the platform interrogates structured experimental data directly, then returns answers scored by evidence and traceable to the exact experiment that produced them. That combination - reach plus receipts - is the product.
Under the hood sits a proprietary technology the company calls the Focal Graph, paired with large language models. Together they let a researcher search multi-omics and chemical biology datasets in seconds, surfacing hidden connections between targets, compounds, diseases and pathways that no single database would reveal on its own.
The mission the team writes on the wall is unusually plain for a software company: to increase the world's ability to discover cures.
Scientific intelligence, not just search.
Modern biology does not suffer from a shortage of data. It suffers from findability. A single disease target might be referenced across genomics screens, proteomics runs, chemical assays and clinical records held in incompatible formats, most of it never read together. A scientist can spend weeks manually stitching those threads into a coherent picture - and still miss the connection that mattered.
Plex's bet is that the bottleneck is not the volume of experiments but the difficulty of asking them a question. By loading minimally processed data - rather than tidy, pre-digested summaries - the platform preserves the messy signal where unexpected relationships hide. Then it makes that signal answerable.
The stakes are financial as much as scientific. A drug program can consume enormous capital before anyone learns a target was a dead end. Plex's target evaluation scores candidates across multiple evidence dimensions before that money is committed, letting teams fail cheap and early, then move quickly on what survives scrutiny.
And because every answer links back to source data, the platform addresses a quieter problem: trust. In R&D, a black box that cannot show its work is close to useless when a single decision costs millions.
Billions of minimally processed data points from multi-omics, chemistry and clinical sources load into one graph.
The Focal Graph links targets, compounds, diseases and pathways into a navigable web.
LLMs turn a scientist's plain question into a search across the graph, answered in seconds.
Every result is evidence-scored and links back to the exact experiment behind it.
The scientific intelligence engine - a biomedical knowledge graph made searchable with evidence-backed answers.
Knowledge-graph technology connecting targets, compounds, diseases and pathways, paired with LLMs for transparent AI.
Integrates multi-omics evidence to surface novel, druggable disease targets.
Multi-dimensional scoring across seven evidence dimensions to de-risk a target before capital commitment.
Identifies predictive and prognostic biomarkers from integrated experimental data.
An agent-based AI that autonomously plans and executes multi-step research campaigns - and reports the strategy and validation data it used.
Programmatic access to the biomedical knowledge graph for custom queries and integrations.
Confidential, secure integration of a client's internal experimental data into their private Plex graph.
Give AutoPlex a research objective and it decomposes the problem, runs the searches, reads the results, and iterates - a campaign that would take a team weeks, run on its own.
Plex serves drug discovery scientists, computational biologists, medicinal chemists and R&D executives at more than 40 pharmaceutical and biotech organizations, alongside academic and translational science groups. Publicly referenced users and logos include:
Business model
Plex operates as B2B enterprise software and discovery services. It licenses access to its platform and knowledge graph, offers enterprise tiers that securely fold in a client's proprietary data, provides API access for developers, and runs consulting-style discovery engagements.
// Illustrative emphasis across Plex's published use cases - relative, not audited figures.
The market for AI in drug discovery is crowded - BenevolentAI, Recursion, Insilico Medicine, Causaly, Ontoforce, Euretos and BenchSci all court the same scientists, as do the in-house bioinformatics teams at large pharma. Plex draws two clear lines against them.
The first is data depth. Many tools reason over summarized literature; Plex reasons over the minimally processed experimental data underneath it. That choice preserves signal that summaries flatten away, which is where novel connections tend to live.
The second is transparency. The Focal Graph is built so that every answer is traceable to its source evidence and scored, rather than delivered as an unexplained recommendation. For teams making multimillion-dollar bets, an AI that can defend its reasoning is worth more than one that is merely confident.
Together those choices position Plex less as a chatbot for biology and more as an evidence layer - a place where a scientific claim and its proof arrive in the same answer.
A three-time founder with 20+ years across life sciences, cloud and AI. Held senior roles at Vertex Pharmaceuticals, Dotmatics and PerkinElmer.
Early microarray pioneer with a Ph.D. from George Church's lab at Harvard. Spent 14 years at Novartis before founding Plex in 2017.
MIT-trained engineer with cloud-security depth. Previously led architecture at CloudLock, acquired by Cisco for $300M in 2016.
Behind the leadership sits a bench of principal scientists drawn from Rockefeller University, Johnson & Johnson, Celgene and the Cleveland Clinic - a team where drug-discovery domain knowledge and software engineering share the same room.
Douglas Selinger establishes Plex Research to apply search-based AI to biomedical experimental data.
First Star Ventures, Boston Seed Capital and others back the early platform.
The company closes a Series A round led by First Star Ventures on December 30, 2020.
Adoption passes 40 pharma and biotech teams, including Pfizer, Merck, Lilly and Takeda.
Plex launches its autonomous AutoPlex agent and partners with Ginkgo Bioworks on compound mechanisms of action.
Plex sits at the junction of three currents: the explosion of multi-omics data, the maturing of knowledge-graph infrastructure, and the arrival of large language models capable of translating a human question into a machine query. Few companies operate credibly across all three; Plex's lineage - a founder from a leading genetics lab, an architect from cloud security, scientists from big pharma - is built precisely for that overlap.
Its footprint is deliberately focused. This is not a general-purpose research assistant but an evidence layer for the specific, high-stakes work of finding and vetting drug targets. That focus is also its leverage: a nine-person core team supporting 40+ drug programs is only possible when the product does one hard thing well.
As autonomous agents like AutoPlex mature, the company's position shifts from tool to collaborator - software that does not just answer the question a scientist types, but proposes and pursues the next one.
It provides an AI-powered scientific intelligence platform that lets drug discovery scientists query billions of biomedical experimental data points and get evidence-backed answers traceable to source data.
Instead of summarizing published papers, Plex queries structured, minimally processed experimental data - genomics, proteomics, compound activity and clinical data - via its Focal Graph, exposing connections that summarized literature hides.
AutoPlex is Plex's agent-based AI system that autonomously plans and executes multi-step drug discovery campaigns - such as target sweeps and safety profiling - and reports the strategy and validation evidence it used.
More than 40 pharmaceutical and biotech teams, including Agios, Eisai, Lilly, Merck, Pfizer, Takeda, Recursion and Ginkgo Bioworks, along with academic research groups.
Plex Research is headquartered at 625 Massachusetts Avenue in Cambridge, Massachusetts. It was founded in 2017 by CSO Douglas Selinger and is led by CEO Brian Gilman.