Cyclica

A pharmaceutical executive is staring at a clinical trial failure report. Five years of development. $200 million invested. Thousands of patients enrolled. The drug showed vigorous preclinical activity against its target protein, clean animal toxicology, and a well-understood mechanism of action. Then Phase II revealed unexpected liver toxicity — and the program was terminated immediately. Post-hoc analysis identified the cause: the compound was hitting an off-target protein in hepatocytes that single-target screening had never evaluated. It is a scenario that repeats itself across the pharmaceutical industry, accounting for a significant share of the 90% of drug candidates that fail to reach approval. Cyclica, the Toronto-based AI drug discovery company founded in 2013, has earned a 2026 Global Recognition Award for building the technology that prevents this exact failure — a deep learning platform that screens small molecules against the entire human proteome in less than one second, achieving a 64% actionable hit rate in pharmaceutical partnership programs, three to six times the industry norm.

Technical Innovation and Architecture

Classical computational drug discovery operates on a single-target paradigm: optimize a compound’s potency against the intended receptor, run ADMET models, and hope the rest of the proteome does not interfere. This approach ignores a fundamental biological reality — every small-molecule drug binds to multiple proteins simultaneously. Some of those off-target interactions destroy clinical programs. Others create therapeutic opportunities that standard screening would never surface. MatchMaker™, Cyclica‘s deep learning engine, addresses this directly. Trained on millions of drug-target interaction records and thousands of protein 3D structures, MatchMaker performs a proteome-wide evaluation — predicting how a compound interacts with approximately 20,000 human proteins — in under one second. What would require days to weeks of mass spectrometry experiments, thousands of dollars, and access to a limited protein panel is compressed into a computational query with near-zero marginal cost. The architecture is deliberately hybrid: deep learning pattern recognition operating on top of first-principles computational biophysics, because Cyclica’s leadership consistently argued that pure machine learning without biophysical grounding fails precisely where pharmaceutical decisions are most consequential.

MatchMaker’s outputs feed into two integrated platforms. Ligand Express™ maps a compound’s complete mechanism of action, identifies off-target interaction risks before synthesis, and enables systematic drug repurposing by revealing new therapeutic indications for existing molecules. Ligand Design™ performs large-scale exploration of chemical space, generating novel compound candidates optimized against multiple simultaneous biological constraints. A third engine — POEM (Pareto-Optimal Embedded Modelling) — builds ADMET toxicity models using a parameter-free supervised learning approach, predicting how a candidate will behave in the human body before a single milligram is synthesized. External validation arrived in January 2021, when a collaboration with AstraZeneca demonstrated that MatchMaker outperformed mass spectrometry — the experimental gold standard — for correctly identifying small-molecule off-target interactions at a fraction of the time and cost.

Market Strategy and Leadership

Cyclica’s origin story is precise and credible. In 2011, bioinformatics scientist Jason Mitakidis presented the proteome-wide AI screening concept at the University of Toronto’s Rotman School of Management MBA case competition and won outright. Naheed Kurji — then an MBA student with a background in biology and two years of humanitarian fieldwork across South and Central Asia and East Africa — saw both the scientific validity and the commercial case. By 2013, they had co-founded Cyclica, and by April 2016, Kurji had transitioned from CFO to President and CEO, overcoming early investor skepticism about a non-PhD executive by building a portfolio of over 100 global pharmaceutical and biotech partnerships. Kurji’s consistent public position was notable for its candor: he framed AI as a powerful, scope-defined tool rather than a universal solution, calling out data quality and bias management as the sector’s primary ethical responsibilities.

The company raised CA$37.5 million in total institutional funding, including a CA$23 million Series B led by Drive Capital in June 2020. With approximately 60 employees managing 50 active programs simultaneously, Cyclica demonstrated that a focused AI team with a validated, scalable platform could serve the global pharmaceutical industry without growing into a large-scale clinical organization. In May 2023, Recursion Pharmaceuticals (NASDAQ: RXRX) acquired Cyclica for $40 million USD, integrating MatchMaker into RecursionOS alongside Recursion’s automated wet labs capable of millions of experiments weekly and one of biopharma’s most powerful supercomputing systems. Recursion CEO Chris Gibson described the combination as giving Recursion “the most complete, technology-enabled drug discovery solution in the biopharma industry” — a claim reinforced when Recursion subsequently acquired Exscientia for $688 million in August 2024.

Industry Impact and Future Vision

The 64% actionable hit rate Cyclica achieved across partnership programs is the platform’s most direct measure of industry impact — and its most important number. The pharmaceutical industry’s typical preclinical success rate sits at 10–20%. Cyclica’s programs ran at three to six times that rate because the platform was identifying biologically valid targets and filtering out toxic liabilities before any physical validation work began. For pharmaceutical partners, this translates directly into fewer terminated programs, reduced synthesis costs, faster progression to meaningful candidates, and lower clinical trial failure rates driven by unforeseen toxicity. Over 100 global pharmaceutical and biotech organizations made that calculation and chose to partner with Cyclica — including AstraZeneca, Merck KGaA, Samjin Pharmaceutical, and PrecisionLife.

Beyond commercial partnerships, Cyclica ran the Academic Partnership Program (CAPP), extending platform access to universities in Central Asia, South America, and Africa without financial benefit—a structural equity commitment to broaden AI drug discovery capabilities beyond well-resourced pharmaceutical markets. A Bill & Melinda Gates Foundation grant funded work on neglected tropical diseases, and co-founding membership in the Alliance for Artificial Intelligence in Healthcare (AAIH) positioned Cyclica at the center of industry-wide conversations about ethical standards for AI in drug development. Within RecursionOS, Cyclica’s MatchMaker technology now operates at a scale it could not reach independently — millions of weekly wet-lab experiments validating proteome-wide computational predictions in a closed feedback loop that improves with every cycle. The 2026 Global Recognition Award recognizes Cyclica for proving what many in the pharmaceutical industry doubted was achievable: that a small team armed with a rigorously built AI platform could predict what the entire human proteome does with a drug molecule — fast enough, accurately enough, and affordably enough to change how medicines are discovered.