Oncology now generates vast, heterogeneous multi-omic data that must be synthesized in real time to support precision medicine and clinical-trial readiness. Expanding on the broader concepts of applied ML in cancer care, this session presents a practical model for AI-enabled multidisciplinary oncology using SYNERGY-AI OS and Reticulum Nexus—multi-agent systems that integrate diagnostics, imaging, genomics, pathology, and clinical context. Examples illustrate improved biomarker identification, use of digital pathology, radiomics, eligibility simulation, decision support, and solutions to the trial-enrollment paradox. Attendees will gain a deployment-ready blueprint for scalable, explainable, and equitable AI in cancer care.

AI is reshaping the full lifecycle of digital diagnostics from biomarker discovery through development, regulatory approval, commercialization, and clinical adoption. Foundation models accelerate progress by improving accuracy at the outset and reducing data and validation requirements. These advances, combined with close collaboration across pathology, biopharma, and diagnostic organizations, enable scalable and trusted clinical solutions that enhance the effectiveness of pathologists and streamline laboratory workflows.

It seems the field has been 1-2 years away from digital pathology taking hold for the past 15 years. However, now is different. As seen with NGS and liquid biopsy, it isn't until pharma includes new technologies in clinical trials that we see a true inflection point in adoption. This presentation will highlight how the field is changing to support this paradigm shift.

Agentic AI offers a new pathway for integrating multi-omic biomarkers, clinical guidelines, FDA label foundations, and real-world evidence into clinically reliable decision-support systems. This session will highlight emerging methods for unifying genomic, transcriptomic, and metabolomic signals with authoritative medical knowledge, enabling transparent, reproducible, and guideline-aligned therapeutic recommendations. Key technical considerations for scalability, interpretability, and privacy-preserving computation will be discussed, along with practical applications across diagnostics, clinical research, and health-system deployments in diverse global settings.

Vikram Chaudhery will discuss his experience helping healthcare companies develop new technologies, from early ideas to products used by doctors and patients. He will highlight what works, what doesn’t, and the practical steps needed to bring diagnostic tools and AI applications into everyday care. Attendees will leave with a grounded understanding of the challenges and opportunities in developing and using new precision medicine tools.

Long-running AI agents are transforming software engineering and other domains where outputs are verifiable: code compiles, tests pass, proofs check. Scientific discovery offers no such luxury: ground truth is, by definition, unknown. With over twenty AI "co-scientist" systems now deployed across biology and chemistry, how do we know they work? This talk surveys the emerging landscape of scientific agent evaluation, drawing from the tradition of critical assessments in biology community-driven challenges like CASP that decoupled method development from evaluation and, in doing so, gave us AlphaFold and discusses what extending this paradigm to autonomous scientific agents could look like.

The realization of precision medicine in the multiomics age is slower than initially envisioned, primarily due to our still-incomplete knowledge of the complex biological circuitry underlying human health and disease. This is changing, thanks to the convergence of computational power, massive multimodal datasets, and causal AI, resulting in Gemini Digital Twins that accurately model any disease and accelerate precision medicine discovery and development (neurodegeneration and oncology examples will be presented).

Whole-genome-derived PRS and AI-driven risk models are transforming proactive care. The session reviews advances in accuracy, population-health applications, and clinical workflows that enable earlier detection and targeted interventions. The focus is on how these innovations reduce downstream costs, improve outcomes, and support large-scale preventive health strategies across diverse patient populations.

Element Biosciences’ AVITI24 enables true single-cell multi-omic profiling, capturing transcriptomic, proteomic, and morphological data from the same cell. This integrated view provides deep insight into cellular machinery and responses to perturbations at pathway-level resolution. We present AI-driven approaches that leverage pseudo-bulk representations of single-cell data to reduce noise while preserving biological signal. By mapping pathway activation and deactivation across perturbations, this framework enables systematic profiling and prediction of combinatorial effects, accelerating target identification and drug discovery.

We will explore how accelerated and integrated compute is reshaping the landscape of biobank-scale datasets, weaving together genomics, imaging, and phenotypic data into validatable models. Expect a whirlwind tour through: 1) Ultra-fast sequence alignment, variant calling, single cell data processing; 2) Real-time genomic discretization, estimating cis/trans effects on variant effects via haploblock architecture and feeding data into models; 3) Accelerated tuning and running of genomic, protein structure, and imaging models; 4) Open-source approaches to federated learning, particularly around biobanks; 5) Accelerated data science, including vector search, large graphs and XGBoost; 6) Knowledge graphs as dynamic memory systems (GNN - LLM co-embedding). We'll close by tackling the unglamorous but essential bits: validation, contextualization, and the digital hygiene required to keep model-generated data from becoming biomedical junk. Think of it as a roadmap toward smarter, faster, and more trustworthy data-driven healthcare.

This presentation will discuss how cutting-edge biomarker discovery and AI modeling are reshaping precision medicine. The discussion will highlight emerging methods that integrate dynamic biomarker signatures with adaptive machine-learning systems to support real-time clinical decision-making. Attendees will gain insight into how next-generation AI can accelerate diagnostics, personalize therapy, and strengthen the resilience of healthcare delivery.

Functional Precision Medicine (FPM) redefines precision oncology by testing live tumor cells against hundreds of FDA-approved drugs to guide individualized therapy. This session explores how integrating functional data with genomic insights enhances predictive accuracy, accelerates clinical decision-making, and bridges translational gaps. Real-world clinical results demonstrate FPM’s potential to expand precision medicine access, improve outcomes, and inform regulatory and reimbursement pathways for broader implementation.

Sygnomics delivers personalized oncology care by combining Institute for Systems Biology (ISB) patented network biology with explainable AI. The SPOT-AI agentic platform offers a white-labeled SaaS concierge for all patients, their providers, and pharma. Currently targeting glioblastoma and multiple myeloma, Sygnomics is expanding its "pan-cancer" model to include lung, pancreatic, colorectal, breast, and liver cancers, using mechanistic insights to more accurately assess prognosis and optimize therapy selection.

We present a modular AI framework that transforms real-world oncology data into predictive insights using clinical, genomic, and image-based inputs. The platform supports drug response prediction, immune therapy stratification, and transcriptomic reconstruction from routine diagnostics. Validated across enterprise-scale real-world cohorts, these models uncover clinically relevant insights that enable precision medicine development and clinical utility.

Precision medicine is fragmenting disease into molecular subtypes too small for traditional evidence frameworks, leaving many patients and clinicians without guidance. When trials become impossible and the evidence base breaks, AI offers a new paradigm by integrating diagnostics, mechanistic biology, and real-world signals to generate individualized, continuously updated insight. My own rare cancer case illustrated this shift. Medicine must now evolve from static, cohort-based proof to dynamic, patient-level learning.

Genialis is building an AI therapeutic recommendation engine for every cancer patient. We call this engine the “Genialis Supermodel,” comprised of a transcriptomics Foundation Model and a comprehensive library of machine learning algorithms that cover all therapeutically relevant cancer biology. The Supermodel yields a new class of biomarkers to match patients with promising new therapies such as KRAS inhibitors, immune agents and all other targets/drugs.