A biomarker consists of two components – one or more "features" and an algorithm that uses those features to provide a diagnostic or prognostic classification. Advanced computational methods including artificial intelligence and machine learning, provide unique opportunities to integrate features from diverse datasets, provided we have the right data on which to train our models, the foresight to avoid classification pitfalls, and a commitment to reproducible research.

 T cell–inflamed gene expression profile (GEP) and tumor mutational burden (TMB) are clinically validated biomarkers that independently predict pembrolizumab response. KEYNOTE-495/KeyImPaCT is a group-sequential, adaptively randomized, multisite, open-label, Phase II study investigating first-line three pembrolizumab-based combinations in patients with advanced NSCLC. Recent study results demonstrate the feasibility and clinical usefulness of prospective GEP and TMB assessment to study the clinical activity of three pembrolizumab-based combination therapies in 1L patients with advanced NSCLC. Although sample sizes were small, the GEPhigh TMBhigh subgroup demonstrated the best response among the biomarker subgroups for all three combination therapies.

The pace of new data acquisition from the diverse assays to measure biomarkers in oncology is accelerating. These diverse assays and the complementary nature of the data generated make possible the realization of Precision Medicine. These range from new imaging molecular probes to measure key molecular targets; to digital pathology; to many analytes from liquid biopsies; to data analysis from AI, machine learning and validation from large, annotated and standardized databases.

This session will introduce the Genomics Organization for Academic Laboratories (GOAL), a consortium effort dedicated to the advancement of genomic testing at academic and non-profit laboratories. The session will begin by covering common obstacles to successful development of NGS assays, opportunities for inter-laboratory cooperation, and current and future initiatives of this multi-institutional molecular pathology collective.

While the current genome-centric approach to Precision Oncology has extended the lives of subsets of patients, many patients do not respond to the selected therapy, and many whose tumors initially respond have a high chance of recurrence as resistant disease. New approaches are required to capture complex clinical phenotypes and better match patients to efficacious therapies. Novel strategies will be discussed to enable NextGen diagnostics.

FFPE tissue samples with a low abundance of nucleic acid continue to present a challenge for cancer genomics. Such samples frequently fail to produce sufficient DNA or RNA for downstream analyses, which can impact patient care or reduce the statistical power of a study. Purigen’s automated Ionic Purification System uses isotachophoresis to increase quantity and quality of nucleic acids extracted from FFPE and requires 75% less hands-on time than column- or bead-based methods. In this workshop, we will review data that demonstrates the ability of the Ionic system to recover increased yields of high-quality DNA, RNA, or miRNA from FFPE tissue samples that prove challenging for conventional methods.

Translating advances in our knowledge of human tumor biology into optimized clinical treatments is a major challenge because cancer progression and tumor drug response is unique to every patient.  We will show how functional profiling in vivo tumor response to 20 different drugs simultaneously addresses this unmet medical need by the discovery of novel and potent immuno-oncology combination therapies and clinically predict standard-of-care oncology drug response in glioblastoma patients.

.

The first companion diagnostic for oncology precision medicine was HER2 IHC over 20 years ago. IHC is semi-quantitative and poor at measuring spatial heterogeneity. We have developed a deep learning computational pathology method that scores HER2 and other IHC markers at low levels, quantitively of each cancer cell and allows scoring of spatial heterogeneity among positive and negative cells. Application of this method to better predict response to ADCs will be presented.

Herein we describe an approach to use machine learning to predict oncogenic drivers (BRAF mutations and NTRK-gene fusions) on digital pathology images in thyroid tumors. Bayer-developed deep learning models represent a promising approach, based on universally available histology slides, in a tumor indication where NGS is not routinely used. The current approach uses rich and diverse datasets for training and evaluation as well as methods for domain-agnostic learning.

Although Precision Medicine therapies have demonstrated benefit to cancer patients, if biomarker testing is not performed, they may receive suboptimal care. Four key barriers to the use of biomarker diagnostics have been identified, and examples of strategies to overcome these barriers will be discussed with respect to liquid biopsy tests. Intrinsic characteristics of liquid biopsy tests make them well-suited to overcome some barriers to use in today’s healthcare system but also at a disadvantage to other barriers.

Alex’s presentation will focus on the key features of different cancer biomarkers and their correlative analysis for IO therapeutics and related combinations.

The Partnership for Accelerating Cancer Therapies (PACT), a Cancer Moonshot public-private partnership, overseen by FNIH, involving NCI, FDA, pharmaceutical companies, non-profits, patient advocates, and the Cancer Immune Monitoring and Analysis Centers and Cancer Immunologic Data Commons (CIMAC-CIDC) Network, provides a systematic approach to immuno-oncology biomarker investigation developing standardized/harmonized biomarkers and assays and implementing them in clinical trials. This presentation describes network establishment and summarizes PACT-CIMAC-CIDC Network progress in assay harmonization to enable cross-trial and cross-laboratory data analysis.

In-depth analyses of functional T cell responses and circulating tumor DNA (ctDNA) following individualized immunotherapy provide critical insight into identifying patients with the potential for durable disease control. While ctDNA assessments can be scaled up for larger studies, the low-throughput nature of functional assays necessitates exploration of alternative assays measuring other T cell metrics, such as TCR sequencing, as a proxy for functional responses.