Panakeia's AI Software Achieves Landmark Validation for Rapid Colorectal Cancer Molecular Profiling
核心洞察
Panakeia (搜索)'s PANProfiler Colorectal (搜索) (MSI (搜索)/MMR (搜索)) software demonstrated 93.83% agreement with standard laboratory testing in the largest real-world, multi-site validation study of AI molecular profiling to date.
The AI software successfully returned results on 86.55% of cases across 1,243 colorectal cancer (搜索) patients from three UK healthcare institutions, delivering molecular biomarker analysis in minutes rather than weeks.
Published in npj Digital Medicine, the study addresses critical gaps in molecular testing access, with national audits showing substantial proportions of eligible patients experience testing delays of several weeks.
Panakeia (搜索) has published landmark clinical validation results demonstrating that artificial intelligence can deliver rapid molecular profiling for colorectal cancer (搜索) patients directly from routine diagnostic images. The study, published in npj Digital Medicine and led by the University of Leeds and Leeds Teaching Hospitals NHS Trust, represents the largest real-world, blinded, multi-site validation to date of AI-enabled molecular profiling from routine diagnostic images.
The company's PANProfiler Colorectal (搜索) (MSI (搜索)/MMR (搜索)) software achieved 93.83% overall agreement with standard laboratory testing for microsatellite instability (搜索) (MSI) and mismatch repair deficiency (搜索) (MMR) - molecular biomarkers that directly inform treatment decision-making for colorectal cancer (搜索) patients. The software correctly identified 92.54% of cases where the biomarker was present (MSI-H/dMMR) and 94.02% of cases where it was absent (MSI-stable).
Breakthrough Performance in Real-World Clinical Settings
The blinded, multi-centered evaluation included 3,576 images of routine tissue samples from 1,243 colorectal cancer (搜索) patients across three independent healthcare institutions in the United Kingdom: University of Leeds/Leeds Teaching Hospitals NHS Trust, Queen's University Belfast/Belfast Health and Social Care Trust, and the University of St Andrews.
PANProfiler Colorectal (搜索) (MSI (搜索)/MMR (搜索)) successfully returned results on 86.55% of cases, representing the highest test replacement rate achieved to date among AI solutions for molecular profiling from tissue images. The study was conducted under rigorous blinded conditions across sites with differing patient populations, diagnostic protocols, tissue preparation methods, and scanner types, reflecting the variability encountered in routine clinical practice.
Dr Nic Orsi, Associate Professor in Histopathology and Industrial Innovation at St James's University Hospital and Leeds Teaching Hospitals NHS Trust, said: "This study was designed to rigorously and independently assess whether AI-based MSI (搜索)/MMR (搜索) testing can perform reliably across real-world clinical data from multiple centres. By analysing thousands of cases under blinded conditions, we were able to demonstrate that PANProfiler Colorectal (搜索) (MSI/MMR) delivers consistent performance comparable to standard laboratory testing, using only routine histology images."
Addressing Critical Healthcare Gaps
Laboratory-based molecular profiling is standard of care for patients diagnosed with colorectal cancer (搜索), as mismatch repair status directly informs treatment decision-making, including immunotherapy eligibility, inherited cancer risk assessment, and prognosis. However, significant gaps persist between guideline recommendations and real-world practice.
National audits and patient surveys show that a substantial proportion of eligible patients do not receive molecular biomarker testing, or experience delays of several weeks due to laboratory capacity constraints and fragmented workflows. During these delays, clinical decisions may be made with incomplete molecular information.
PANProfiler Colorectal (搜索) (MSI (搜索)/MMR (搜索)) addresses these challenges by delivering mismatch repair deficiency (搜索) status directly from routine brightfield images already generated during standard diagnostic processes. By providing results in minutes rather than weeks, the software supports faster, more informed clinical decision-making and enables earlier triage for confirmatory testing where required.
Platform Validation Across Cancer Types
The colorectal cancer (搜索) validation builds on prior platform-level evidence for Panakeia (搜索)'s PANProfiler technology. In 2024, Panakeia published a study in Nature Communications Medicine, recognized as one of the journal's top 25 papers of the year, demonstrating the platform's ability to identify thousands of biomarkers across more than 30 cancer indications from routine diagnostic images.
Independent validation of PANProfiler Breast, also conducted in collaboration with the University of Leeds and published in Clinical Breast Cancer (搜索), further supports the generalizability of the approach across tumor types and clinical contexts.
Regulatory Recognition and Clinical Deployment
PANProfiler Colorectal (搜索) (MSI (搜索)/MMR (搜索)) has been selected for the UK Medicines and Healthcare products Regulatory Agency's (MHRA) AI Airlock programme, which supports real-world evaluation of AI medical devices. Panakeia (搜索) has also participated in the Friends of Cancer Research Digital PATH Project, a multi-stakeholder initiative examining agreement and reliability across AI-based digital pathology platforms.
The software is UKCA-marked and deployed in clinical sites across the UK, with international deployment plans advancing rapidly. Pahini Pandya, Founder and Chief Executive Officer of Panakeia (搜索), commented: "This npj Digital Medicine publication validates our vision to make precision medicine accessible to patients and doctors globally. Our biology-first approach has enabled us to build software that delivers robust molecular insights from data already generated in care."
Technology and Implementation
PANProfiler Colorectal (搜索) (MSI (搜索)/MMR (搜索)) is an AI software that determines mismatch repair deficiency (搜索) status directly from H&E-stained images of colorectal cancer (搜索) tissue samples, entirely digitally. The software integrates into existing hospital workflows by leveraging data already generated during standard of care processes, without requiring additional laboratory testing or changes to existing hospital workflows.
The PANProfiler platform uses a biology-first AI approach trained on petabytes of population-scale data to determine multi-omic biomarker status across DNA, RNA, proteins, and metabolites directly from routine brightfield images of cells and tissues. The platform can profile thousands of biomarkers in more than 30 cancer types, supporting drug discovery, clinical development, and patient care.