Seeing What Genes Can’t: Acurion’s AI-Powered Phenotypic Detection Delivers Immediate, Actionable Cancer Insights

Today, there are more than 100 precision therapies approved for cancer treatment, yet fewer than one in twenty patients receive them. San Diego–based Acurion is working to change that, reshaping precision oncology by shifting the focus from genotypic testing to phenotypic analysis.

Its flagship platform, OncoGaze™, combines advances in microscopy imaging and artificial intelligence to read tumor biology directly from routine pathology slides. Rather than inferring what might be happening from DNA alone, OncoGaze™ detects the effects directly in tissue—delivering clinically actionable biomarkers in minutes instead of weeks. Validated in more than 1,400 patients and published in the Journal of Clinical Oncology, it is the first AI solution of its kind with peer-reviewed clinical outcome data.

Currently, OncoGaze™ identifies homologous recombination deficiency (HRD) in breast and ovarian cancer, enabling oncologists to find up to three times more patients eligible for life-extending therapies. Acurion is now expanding its reach, with a first-of-its-kind agreement with the Pancreatic Cancer Action Network to access the world’s largest pancreatic cancer dataset.

Behind this breakthrough is Dr. Ludmil Alexandrov—one of the world’s most highly cited computational biologists—whose pioneering work on mutational signatures has transformed cancer diagnostics. His contributions have been recognized with the Science Prize for Young Scientists, the Harold M. Weintraub Award, the Carcinogenesis Young Investigator Award, and Forbes’ 30 Under 30 in Science and Healthcare.

OncoGaze™ represents a major departure from traditional genetic testing by detecting biomarkers directly from pathology slides. What inspired this approach, and why do you believe it is a game-changer for precision oncology?

OncoGaze™ grew out of a simple but powerful observation: every tumor’s genome leaves a footprint in its histology. My lab has spent the last decade decoding molecular signatures in DNA, and we asked whether those same biological processes might be visible in pathology slides. The result was striking: with AI, we can identify homologous recombination deficiency (HRD) directly from H&E images. This approach bypasses the cost, complexity, and turnaround time of sequencing, making precision oncology faster and more widely accessible.

Homologous recombination deficiency (HRD) is a key biomarker for guiding cancer therapy, yet it’s often under-detected with current methods. How does OncoGaze™’s ability to identify up to three times more HRD-positive patients change the treatment landscape for those cancers?

Current genomic assays miss a substantial fraction of HRD-positive tumors, which means patients who could benefit from platinum chemotherapy or PARP inhibitors often don’t receive them. OncoGaze™ identifies up to three times more HRD-positive cases compared to standard molecular tests. This has profound implications: more patients guided to effective therapies, fewer missed opportunities, and improved survival outcomes in ovarian, breast, and soon pancreatic cancer.

Your background in computational biology and genomics is extensive—how did your previous scientific research inform the development of OncoGaze™, and what gaps in cancer diagnostics were you most determined to close?

My academic work on mutational signatures showed that DNA damage processes shape the cancer genome in predictable ways. Mutational signatures are the characteristic patterns of mutations left behind by different processes—such as defective DNA repair or exposure to carcinogens—that act like fingerprints in the tumor genome. That research revealed gaps: diagnostics relying purely on genetics were too narrow, slow, and expensive. OncoGaze™ bridges that gap by translating years of genomics insight into an image-based tool. The narrative is simple: every H&E slide is not just morphology, but a histology record that contains clinically useful information waiting to be read.

Many AI tools in medicine face challenges with reproducibility and clinical adoption. What steps did Acurion take to ensure that OncoGaze™ is both scientifically rigorous and practical in real-world oncology settings?

From the outset, Acurion committed to scientific rigor—peer-reviewed evidence in JCO, validation across thousands of patients, and collaborations with leading cancer centers. Just as important, we designed OncoGaze™ to slot seamlessly into existing pathology workflows: no new tissue prep, no extra equipment, just digitize the H&E slide and run the analysis. That combination—robust science plus clinical practicality—will be the key to adoption.

Unlike genetic testing, which can take weeks, OncoGaze™ delivers instant results. How do you see this speed transforming the patient journey—from diagnosis to treatment decisions?

Genomic assays can take weeks, delaying critical treatment decisions in breast, ovarian, and especially pancreatic cancers where every day counts. OncoGaze™ changes that equation by delivering results in minutes. Instead of patients waiting in uncertainty, their oncologists know HRD status right away and can start the right therapy immediately. This speed transforms the patient journey—turning weeks of anxious delay into same-day, confident treatment decisions.

AI in healthcare sometimes raises concerns around transparency. How does OncoGaze™ ensure explainability in its outputs so that oncologists and pathologists can confidently integrate the results into their clinical decisions?

OncoGaze™ was built as a true white-box AI. Rather than acting as a black box, it highlights the precise regions and features within an H&E slide that drive its predictions, generating heatmaps that pathologists can directly interpret. Even more importantly, these highlighted features match biological expectations—capturing the patterns that arise when DNA repair is defective and HRD is present. The result is an AI that doesn’t just give an answer, it shows why—bridging advanced computation with biological insight. This transparency empowers oncologists and pathologists to trust the output, validate it against their own expertise, and confidently integrate it into clinical care.

While HRD is the first application of OncoGaze™, the platform is designed to be flexible. What other biomarkers or cancer types do you envision it being trained to detect in the near future?

While HRD is the first application of OncoGaze™, the platform was intentionally designed to be flexible and clinically actionable. We are actively training it to identify mismatch repair deficiency (MSI) and tumor mutational burden (TMB), but with a crucial distinction: not simply flagging MSI or TMB, but differentiating the cases that actually confer sensitivity to immune checkpoint inhibitors from those that do not. By focusing on these clinically meaningful signals rather than the biomarker label alone, OncoGaze™ ensures oncologists can make immediate, confident treatment decisions directly from H&E slides.

Beyond oncology, do you see applications for this type of image-based biomarker detection in other disease areas, such as rare diseases or even non-cancerous conditions?

Yes. At its core, OncoGaze™ is a visual biomarker detection engine, and the same principles could eventually extend to rare diseases with distinctive pathology or even non-cancer conditions such as inflammatory disorders where histology reveals underlying molecular states. That said, our current and primary focus is firmly on oncology—where the unmet clinical need is greatest and where rapid, actionable biomarkers can have the most immediate impact on patient survival.

Precision oncology often requires integration of multiple data sources—genomics, pathology, imaging, clinical records. How might OncoGaze™ evolve to fit into a more holistic precision medicine ecosystem?

While it’s true that precision oncology benefits from integrating multiple data types—genomics, imaging, clinical records—the reality is that such integration is usually only partial and largely confined to top cancer centers. Of course, if one has access to all of this information, predictions can be stronger. But realistically, over the next 20 years, this level of integration will only be available to a small fraction of patients worldwide.

OncoGaze™ takes a different approach. We ask: what is the minimum amount of information needed to reliably predict clinical response? Every cancer patient, no matter where they are treated, has an H&E slide from their biopsy. By building our platform on that universal foundation, we ensure broad accessibility while still delivering actionable insights. We can and have integrated additional layers of information, but from a practical perspective, the greatest impact comes from a solution that works from one thing every patient already has.

Looking 5–10 years ahead, what is your vision for how AI-driven diagnostics like OncoGaze™ will reshape not just cancer care, but the healthcare system at large?

Looking 5–10 years ahead, I see AI-driven diagnostics like OncoGaze™ becoming a standard first step in cancer care. Every biopsy slide will be digitized, and AI will be run immediately to deliver actionable results on the spot. That means oncologists won’t need to wait weeks for molecular tests before deciding on therapy—patients can start the right treatment the same day.

This vision is especially powerful outside of elite cancer centers. In many parts of the world, including my home country (Bulgaria), access to advanced sequencing and integrated data is limited or prohibitively expensive. By leveraging the one thing every patient already has—an H&E slide—we can democratize precision oncology, ensuring that fast, informed treatment decisions are available not just in top hospitals but across the globe.