Nature Communications publishes PACpAInt: deep learning decodes pancreatic cancer complexity
Nature Communications today published PACpAInt – a cutting-edge AI model co-developed by Owkin and AP-HP that decodes the complexity of pancreatic cancer, potentially revolutionizing the diagnosis and treatment of a disease with the lowest survival rate of all common cancers.
Led by AI biotech Owkin and Assistance Publique–Hôpitaux de Paris (AP-HP), the largest hospital system in Europe, PACpAInt uses deep learning techniques to analyze and predict different molecular characteristics of pancreatic cancer tumors, including tumor cells and their surrounding environment. PACpAInt leverages self-supervised learning to understand features from over 6 million histology images.
By being able to accurately predict tumor molecular subtypes and patient survival outcomes, and given the expansion of digital pathology worldwide, PACpAInt opens the way for patient stratification based on powerful molecular criteria in routine care and clinical trials.
Researchers used PACpAInt to discover that pancreatic cancers display a high level of intratumor heterogeneity and that nearly one-third of tumors fall halfway between the two main previously described subtypes, a discovery that could impact the creation and targeting of new drugs to tackle the disease. It also identified different components of a tumor's surrounding environment, making it easier to group patients for targeted treatment trials.
Unlike many other AI models, PACpAInt is interpretable by design, allowing pathologists to independently verify its findings. PACpAInt was trained on a large multicentric dataset of 424 digital pathology slides and associated transcriptomics data from 202 patients, and validated using several independent groups of samples, including surgical and biopsy specimens.
Without PACpAInt, molecular subtyping requires costly, lengthy, and complex RNA sequencing. With the expansion of digital pathology, PACpAInt reinforces the potential of using histology-based deep learning models to advance precision medicine in a resource-efficient way. This has the potential to open up precision medicine to far more patients globally.
Charlie Saillard, Senior Data Scientist at Owkin, first author, said:
PACpAInt’s ability to decode the complexity of pancreatic cancer is a major advancement in our ability to understand and treat one of the most lethal cancers. By applying AI to digital pathology slides, PACpAInt is a faster, more accurate and more efficient method of understanding a patient's unique disease. This allows doctors to tailor treatments more efficiently to the individual – an important moment in precision medicine.
Jerome Cros, Professor of pathology at Beaujon Hospital AP-HP, senior author, said:
This tool was developed using the unique histological and molecular resources from four AP-HP hospitals (Amboise-Paré, Beaujon, Pitié-Salpétrière, Saint-Antoine) through a unique collaboration between pathologists from AP-HP, bioinformaticians from Inserm and data scientists from OWKIN. It can remotely subtype pancreatic adenocarcinoma in minutes, paving the way for many applications from basic science (study of intra-tumor heterogeneity) to clinical practice (tumor subtyping in clinical trials).
Remy Nicolle, Researcher at Inserm, and co-senior author said:
By transferring molecular subtyping from complex genomic profiling to standard histology, this tool has revolutionized our understanding of pancreatic adenocarcinoma‘s complex cellular architecture, accountable for its exceptional aggressivity. This tool is able to phenotype large series of tumors at high resolution, unlocking the possibility to analyse the intra-tumor heterogeneity at scale and uncovering previously unknown patterns of tumor evolution.
About AP-HP
The leading hospital and university center (CHU) in Europe, the AP-HP and its 38 hospitals are organized into six university hospital groups (AP-HP. Center - University Paris Cité; AP-HP. Sorbonne University; AP-HP North - University Paris Cité; AP-HP. Paris Saclay University; AP-HP. Henri Mondor and AP-HP University Hospitals. Paris Seine-Saint-Denis University Hospitals) and are organized around five universities in the Ile-de-France region. Closely linked to major research organizations, the AP-HP has eight world-class hospital-university institutes (ICM, ICAN, IMAGINE, FOReSIGHT, PROMETHEUS, lnovAND, Re-Connect, THEMA ) and the largest French health data warehouse (EDS).
A major player in applied research and innovation in health, AP-HP holds a portfolio of 650 active patents, its clinician-researchers sign nearly 10,000 scientific publications each year and more than 4,000 research projects are currently under development, all promoters combined. In 2020, AP-HP obtained the Institut Carnot label, which rewards the quality of research partnership: Carnot @ AP -HP offers industrial players solutions in applied and clinical research in the health sector. The AP-HP also created in 2015 the AP-HP Foundation for Research to support biomedical and health research carried out in all of its hospitals. http://www.aphp.fr
About Owkin
Owkin is the first end-to-end AI biotech company on a mission to understand complex biology and ensure every patient gets the right treatment.
We identify precision therapeutics, de-risk and accelerate clinical trials, and develop diagnostics using AI trained on world-class patient data through privacy-enhancing technologies. We merge wet lab experiments with advanced AI techniques to create a powerful feedback loop for accelerated discovery and innovation in oncology, cardiovascular disease, and immunity and inflammation.
Owkin also founded MOSAIC, the world’s largest spatial multi-omics atlas for cancer research across nine cancer indications.
Owkin has raised over $300 million through investments from leading biopharma companies, including Sanofi and BMS, and venture funds like F-Prime, GV and Bpifrance, among others.
