ROCKVILLE, Md. — Artificial intelligence is changing how molecular pathologists analyze complex genetic information, classify cancers and make decisions about diagnostic testing. But how reliable are these tools, and where can they make a meaningful difference in patient care?
Several studies presented at the Association for Molecular Pathology 2026 Annual Meeting & Expo explore the emerging role of AI in molecular diagnostics, including identifying difficult-to-classify cancers, improving the accuracy of AI-generated reports and helping laboratories use genetic testing more efficiently.
Journalists are invited to attend the meeting in person or sign up for online access to all press materials. Details are available at https://amp26.amp.org/media/media-information/.
Below are just a few of the findings that will be shared by leading experts in molecular diagnostics at AMP 2026.
AI model accurately classifies dozens of blood cancer subtypes
Researchers have developed an artificial intelligence model that uses DNA methylation patterns to identify and classify blood cancers. These cancers include numerous subtypes that can be difficult to distinguish using conventional diagnostic approaches, yet accurate classification is essential for determining prognosis and treatment.
The researchers (from ALMA Genomics, the University of Pennsylvania, the Massachusetts Institute of Technology and several other academic and clinical institutions) developed ALMA 3, an AI foundation model trained using a large collection of (more than 110,000) patient samples.
In independent testing, the diagnostic model detected the presence of hematolymphoid tumors with 99.5% sensitivity and achieved a 90.9% positive predictive value when assigning specific cancer subtypes.
The study included 81 subtypes of blood and lymphatic cancers, including several that are particularly difficult to classify using conventional diagnostic methods.
The findings demonstrate the potential for AI-assisted molecular classification to complement conventional pathology, although additional clinical validation is under way.
This work was led by Pedro Orsini, M.D., of ALMA Genomics. Francisco Marchi, Ph.D., the corresponding author, will give an oral presentation from 7–7:15 a.m. on Friday, Nov. 13, and present the poster from 9:15–10:15 a.m. on Saturday, Nov. 14, at the convention center in Seattle. Abstract H071: An Epigenomic Foundation Model for Hierarchical WHO/ICC Classification of Hematolymphoid Tumors
AI platform makes cancer genetics reporting more reliable, accessible
Scientists at the University of Michigan have developed an open-source, AI-assisted platform designed to help molecular pathologists interpret complex genetic sequencing data and generate clinical reports while reducing the risk of unsupported or inaccurate information.
The platform, called Turnkey Precision Oncology, processes DNA and RNA sequencing results, identifies genetic alterations and generates draft reports for pathologist review. Its AI component, TPO-Agent, draws on established cancer genetics databases and scientific literature, checking that its interpretations are supported by cited evidence.
In testing involving 50 genetic alterations across 10 patients, approximately 95% of the AI system's claims were correctly supported by cited evidence, compared with 65% for a conventional AI approach equipped with similar research tools. The broader platform also demonstrated approximately 95% agreement with a clinical laboratory in identifying individual genetic alterations across 28 reports. Its AI-generated reports showed high agreement with expert interpretations of clinically significant alterations.
The findings suggest that AI systems designed around verifiable evidence and expert oversight could help laboratories produce molecular pathology reports more efficiently and make comprehensive cancer genetic reporting more accessible to community laboratories. Further evaluation will help establish how the platform performs in routine clinical practice.
Both studies were led by Marcin Cieslik, Ph.D., of the University of Michigan. Cieslik will present Abstract I040, “Turnkey Precision Oncology an Open, Validated Read-to-Report Platform for Guideline-Concordant, AI-Assisted Molecular Pathology Reporting,” and Kevin Meng-Lin, B.S., a fellow in Cieslik's laboratory, will present Abstract I068, “Measuring the Accuracy, Hallucination, Bias and Guideline Adherence of AI-Assisted Molecular Pathology Reporting,” during a poster session at 9:15 a.m. on Friday, Nov. 13.
AI-assisted review may cut unneeded rapid genetic testing in lung cancer
Researchers at Memorial Sloan Kettering Cancer Center and the Icahn School of Medicine at Mount Sinai investigated whether artificial intelligence could help pathologists determine which lung cancer specimens need rapid genetic testing.
Testing for mutations in the EGFR gene can help identify patients who may benefit from targeted treatments, but performing multiple tests can consume valuable tumor tissue and laboratory resources.
The team incorporated AI analysis of digital pathology slides into a workflow in which six molecular pathology fellows reviewed 80 consecutive lung adenocarcinoma cases. The project used pathology-related AI models, namely TAPFM and EAGLE.
Based on the AI findings, reviewers indicated that rapid EGFR testing could have been deferred in 26 cases, or 33%, without overlooking any of the confirmed EGFR-positive cases in the study.
These preliminary findings suggest that AI could help laboratories prioritize rapid testing while preserving tissue for comprehensive genomic analysis. However, the approach was less reliable for identifying cases that could forgo rapid testing for certain gene fusions, highlighting the need for further refinement and human-in-loop oversight.
Overall, the authors said, this approach provided a safe and cost-effective way to reduce rapid molecular testing, preserve tissue for more comprehensive molecular testing and make better treatment decisions.
This work was led by Nicholas Dcunha, MBBS, of Memorial Sloan Kettering Cancer Center, who will give a presentation about it during a poster session at 9:15 a.m. on Friday, Nov. 13, at the convention center in Seattle. Senior author Chad Vanderbilt, M.D., also of MSKCC, oversaw the work. Abstract I070: Real-Time Human-in-the-Loop Histopathology AI Genomic Test Triage Reduces Rapid EGFR Testing by 33% in Lung Adenocarcinoma
AI detects molecular differences between liver cancers using routine pathology images
Researchers at Beth Israel Deaconess Medical Center and Harvard Medical School explored whether artificial intelligence could distinguish two major forms of liver cancer by analyzing routine microscope-slide images.
Hepatocellular carcinoma and intrahepatic cholangiocarcinoma arise from different types of liver cells and have different treatment approaches, making accurate classification important.
Using an AI model capable of analyzing pathology images without additional disease-specific training, the team examined 50 digitized tissue slides from The Cancer Genome Atlas public dataset. The model detected visual patterns linked to cancer-type-specific genes, especially those characteristics of liver cells (hepatocytes), and RNA sequencing data from the same tumors supported these findings.
The results suggest that AI may be able to extract molecular information from tissue images that are already collected during routine diagnosis, potentially providing another tool to support cancer classification.
This work was led by Avash Das, MBBS, Ph.D., of Beth Israel Deaconess Medical Center/Harvard Medical School, who will give a presentation about it during a poster session 9:15–10:15 a.m. on Friday, Nov. 13, at the convention center in Seattle. Vikram Deshpande, M.D., co-led the work. Abstract I056: Zero-Shot Whole Slide Image Analysis Distinguishes Hepatocellular Carcinoma from Cholangiocarcinoma Using Pathology Language Foundation Models
AI shows promise for predicting mutations from stomach cancer slides
Scientists at Stanford University investigated whether artificial intelligence could predict cancer-associated genetic mutations using standard pathology images, potentially providing another tool to complement molecular testing.
Although genomic sequencing provides valuable information for cancer diagnosis and treatment, it can be expensive and may not be readily available in every clinical setting.
The researchers analyzed digitized tissue slides from 367 patients with gastric cancer, using an AI foundation model to identify image features associated with mutations in four cancer-related genes. The strongest results were for mutations in the PIK3CA and CDH1 genes, with predictive performance scores of 0.80 and 0.77, respectively, on a scale where 0.50 represents chance-level discrimination and 1.0 represents perfect discrimination.
While the findings require validation in additional patient populations, the study suggests that AI analysis of existing pathology slides could eventually help laboratories prioritize molecular testing.
This work was overseen by Kiran Nijjer of Stanford University, who will give a presentation about it during a poster session at 9:15 a.m. on Saturday, Nov. 14, at the convention center in Seattle. Abstract I069: Deep Learning-Based Prediction of Genomic Alterations from Histopathology Using Large-Scale Pre-Trained Representations in Gastric Adenocarcinoma
###
About AMP
The Association for Molecular Pathology was founded in 1995 to provide structure and leadership to the emerging field of molecular diagnostics. AMP’s more than 3,100 members practice various disciplines of molecular diagnostics, including bioinformatics, infectious diseases, inherited conditions and oncology. Our members are pathologists, clinical laboratory directors, basic and translational scientists, technologists and trainees who practice in a variety of settings, including academic and community medical centers, government and industry. Through the efforts of its Board of Directors, Committees, Working Groups and Members, AMP is the primary resource for expertise, education and collaboration in one of the fastest-growing fields in healthcare. AMP members influence policy and regulation on the national and international levels, ultimately serving to advance innovation in the field and protect patient access to high-quality, appropriate testing. For more information, visit www.amp.org.