PGxAI’s Andromeda Outperforms Frontier AI Labs at Clinical Prescribing Guidance

PALO ALTO, CA, UNITED STATES, September 14, 2026 /EINPresswire.com/ -- PGxAI, a precision-medicine company developing AI-powered clinical decision support for safer and more personalised prescribing, today announced the publication of a peer-reviewed study in npj Digital Medicine demonstrating that PGxAI’s Andromeda AI system significantly outperformed GPT-5, Claude Opus 4 and Grok at generating pharmacogenomic prescribing recommendations. PGxAI has raised $3 million in seed funding from ATEM Capital, Inkberry and Najashi Holding.

Andromeda is a HIPAA-compliant and SOC 2-compliant clinical decision support platform that brings together pharmacogenomics, drug–drug interactions, drug–condition interactions and real-world evidence in a single, clinician-reviewable layer. The platform translates complex biological and clinical signals, including genetic variants, co-medications, and patient conditions, into traceable, auditable prescribing guidance at the point of care. Every recommendation links directly to its underlying evidence sources. Treatment decisions remain with the licensed healthcare professional, so Andromeda provides guidance, not autonomous prescribing.

“Medication response is shaped by far more than a single gene. Clinicians must account for interacting genetic variants, metabolic pathways, co-medications and other patient-specific factors while the scientific evidence continues to expand. Andromeda is designed to synthesize that complexity into evidence-linked guidance that can help clinicians select the appropriate medication and dose for each patient.
— Mike Zack, MD, PhD, MPH, Chief Executive Officer and Co-Founder, PGxAI

The scientific foundation of Andromeda was validated in a 2026 study published in npj Digital Medicine.¹ In a blinded feasibility evaluation, PGxAI-Recommender received the highest expert rating and significantly outperformed all tested large language models from leading frontier AI labs on clinical accuracy and adherence to established CPIC prescribing guidelines (adjusted p<0.01 versus all models). The system is designed to identify up to 573 gene–drug pairs and retrieve nearly 20,000 relevant pharmacogenomic publications, with its evidence base continuously updated through an agentic AI system that autonomously retrieves and structures newly published literature.

“Benchmark performance is the first step; clinical value depends on integration, implementation, governance and accountability. — Burns C. Blaxall, PhD, President and Chief Business Officer, PGxAI

The clinical case for pharmacogenomics-led prescribing is well established. A 2023 multicentre study published in The Lancet involving around 7,000 patients found that genotype-guided prescribing reduced clinically relevant adverse drug reactions by approximately 30 percent compared with standard care.² A 2018 peer-reviewed study estimated the annual cost of prescription drug-related morbidity and mortality from non-optimised medication therapy at $528.4 billion in the U.S., equivalent to 16 percent of total healthcare expenditures.³ Despite the evidence, broad implementation has stalled. The obstacle has been the inability to interpret and apply pharmacogenomic data at clinical scale, which Andromeda is designed to solve.

The announcement comes at a moment of accelerating policy attention on prescribing practices. By July 2025, Medicare coverage for pharmacogenomic testing extended nationwide for tests tied to CPIC- or FDA-recognised drug–gene interactions.⁴ In May 2026, the U.S. Department of Health and Human Services launched an action plan to curb psychiatric overprescribing and promote clinically appropriate deprescribing.⁵ Andromeda supports evidence-based medication optimisation across the full treatment lifecycle, from initiation and dose adjustment through to clinically appropriate discontinuation. For example, established CPIC guidance recognises that genetic variation in CYP2D6, CYP2C19 and CYP2B6 can significantly affect the metabolism, dosing, efficacy and tolerability of multiple serotonin reuptake inhibitor antidepressants.⁶

“Deprescribing is not a separate clinical discipline; it is the stop phase of precision prescribing — the right drug, for the right patient, at the right dose and time, including when to discontinue. As federal policy begins to recognise and support supervised deprescribing, evaluation of pharmacogenomics, drug–drug and drug–condition interactions should be integral to the process. It provides objective evidence about drug metabolism, efficacy and tolerability, while specialist AI can translate a rapidly expanding evidence base into guidance that clinicians can review, audit and trust.”
— Allan Gobbs, Executive Chairman and Co-Founder, PGxAI

<END>

Notes to Editors:

References:
1. doi.org/10.1038/s41746-026-02590-w
2. doi.org/10.1016/S0140-6736(22)01841-4
3. doi.org/10.1177/1060028018765159
4. Pharmacogenomics Testing: Local Coverage Determinations across Medicare Administrative Contractors. U.S. Centers for Medicare & Medicaid Services, Medicare Coverage Database (cms.gov/medicare-coverage-database); most recent: National Government Services LCD L39995, effective July 13, 2025.
5. hhs.gov/press-room/hhs-launches-maha-action-plan-curb-psychiatric-overprescribing.html
6. doi.org/10.1002/cpt.2903

Josh Cobb
Jefferson Communications
josh.cobb@jeffersoncommunications.co.uk
Visit us on social media:
LinkedIn