AACR Annual Meeting 2025

Stop by our booth to experience live demos of our newest technologies and discover how we are leveraging AI to advance precision medicine.

Tempus Lens: Quickly find, access, and analyze multimodal de-identified data records and uncover critical insights to accelerate research and innovation. Leverage our AI assistant, Tempus One, to quickly define detailed patient cohorts using natural language processing and transform complex criteria into actionable datasets in seconds. Explore unstructured clinical data and extract nuanced insights. With Lens, the power of Tempus data is at your fingertips.

Tempus Next: Discover Tempus Next, an AI-powered platform that leverages near real-time curated unstructured and structured patient data to help providers identify the right patient for the right next step in their care. With Tempus Next, life sciences commercial teams can access current, comprehensive, and actionable patient journey insights and proactively address care gaps in guideline-directed precision medicine.

Transforming R&D with Generative AI and Real-World Data

Explore the transformative power of artificial intelligence in oncology research and drug development. Hear from industry leaders about leveraging machine learning models to accelerate discovery and unlock greater potential from real-world datasets. We'll highlight cutting-edge applications of generative AI and discuss how these advancements are shaping a new era in precision medicine.

*This Exhibitor Spotlight Theater is a promotional activity and is not approved for continuing education credit. The content of this Exhibitor Spotlight Theater and opinions expressed by presenters are those of the sponsor or presenter and are not of the American Association for Cancer Research® (AACR)

Stratification Based on PRAME Gene Expression Shows Inverse Survival Associations Among Histology Subtypes in a First-Line Non-Small Cell Lung Cancer Real-World Cohort

This study investigates PRAME expression as a prognostic biomarker in non-small cell lung cancer patients treated with immunotherapy. PRAME, an intracellular antigen overexpressed in solid tumors, is a promising target for pan-cancer immunotherapy. This study analyzes 4,519 non-small cell lung cancer patients, using the Tempus xR RNA-seq assay to quantify PRAME expression levels. Based on the median PRAME expression within the NSCLC cohort, patients were stratified into high and low expressor groups, and their real-world survival was assessed. Results showed a bimodal distribution of PRAME expression in both LUAD/LUSC histologies, with a higher prevalence of high expressors in LUSC. LUSC patients had worse rwOS across all treatments (IO, Chemo, IO+chemo) compared to the LUAD patients. High PRAME LUSC expressors showed improved rwOS in the IO+chemo group, while low LUAD expressors had better survival outcomes. This study concludes that PRAME expression serves as a prognostic biomarker in NSCLC, with different implications based on histological context.

Enhancing Clonal Hematopoiesis Variant Detection in Tumor-Normal Matched Sequencing: Using Machine Learning

This study addresses the challenge of accurately identifying clonal hematopoiesis (CH) variants in tumor-normal matched sequencing data. By developing an algorithm that integrates copy number variations and tumor purity, researchers demonstrate the ability of the Tempus xT platform to distinguish CH from germline and artifactual variants. Validation against higher-depth sequencing showed improvements in sensitivity, precision, and specificity for CH detection. This accuracy-driven approach for identifying CH variants can deepen our understanding of variant biology and potentially improve clinical outcomes

Genetic and clinical landscape of NUTM1 structural variants

Within Tempus' multimodal real-world database, researchers identified 59 patients with a primary diagnosis of NUT carcinoma—an aggressive cancer—81% of whom had a confirmed NUTM1 fusion. Notably, there were 106 additional patients who had a NUTM1 fusion without a corresponding initial NUT carcinoma diagnosis, suggesting a potentially significant underdiagnosis rate. The study found a variety of fusion gene partners, with certain cancer types showing enrichment of specific fusions. With a median overall survival of just over 5 months, the findings suggest that certain cancer types with a high enrichment of NUTM1 fusions may benefit from universal next-generation sequencing to ensure accurate diagnosis and potentially improve outcomes for patients with high-risk cancer types.

Genomic and transcriptomic mediators of resistance to antibody-drug conjugates (ADCs) in metastatic breast cancer (MBC): a comprehensive multi-center study

In this study, researchers examined genomic and transcriptomic underpinnings to ADC resistance mechanisms in metastatic breast cancer (MBC). The analysis included RNA and DNA NGS data from patients in the Tempus multimodal real-world database diagnosed with MBC and treated with FDA-approved ADCs. While no significant differences in target antigen alterations or expression were observed in patients with acquired resistance to sacituzumab govitecan (SG) or trastuzumab deruxtecan (T-DXd), a decrease in both ERBB2 gene amplifications and expression was noted post-trastuzumab emtansine (T-DM1) treatment. Additionally, trends toward higher expression of efflux pump genes were noted in patients with acquired resistance to SG and T-DXd. These findings suggest that differences in ADC mechanisms of action may confer distinct acquired and primary resistance patterns, including alterations in efflux pump and target antigen gene expression.

Metabolic and tumor immune cell landscapes are significantly different amongst KRAS mutational variants in non-small cell lung cancer

Analyzing data from 5,925 non-small cell lung cancer (NSCLC) patients with KRAS alterations within Tempus' multimodal real-world database, researchers identified significant differences in lipid metabolism and immune cell proportions associated with specific KRAS variants. The KRAS G12D variant was associated with a less immunogenic microenvironment, characterized by lower tumor mutational burden and fewer CD8 T cells, potentially impacting the efficacy of immunotherapy. These findings highlight the need for further research to understand how specific KRAS variants may influence treatment responses in NSCLC.

ROS1 single nucleotide variants predict favorable survival outcomes on immunotherapy regimens in non-small cell lung cancer

This study explores the potential of using real-world data (RWD) to uncover novel genetic associations that predict response to immune checkpoint inhibitor (ICI) therapy in non-small cell lung cancer (NSCLC) patients. Analyzing two RWD cohorts, one of which leveraged Tempus' multimodal real-world database, researchers identified a significant association between ROS1 single nucleotide variants (SNVs) and favorable progression-free survival (PFS) in patients treated with ICIs, an effect not observed with non-ICI treatments. The findings suggest that ROS1 SNVs may serve as predictive biomarkers for ICI responsiveness in the first-line setting, potentially offering new insights into personalized treatment strategies for NSCLC.

Visium HD combined with deep-learning-based cell segmentation on H&E images yield accurate cell annotation at single-cell resolution

By training a neural network for cell segmentation on H&E stained images, the team successfully mapped 2 𝜇m-resolution Visium HD data to single-cell gene counts in non-small cell lung cancer (NSCLC) samples. The resulting cell clusters, annotated using a large language model (LLM), showed a high degree of concordance with pathologist annotations, accurately identifying lymphocytes, cancer cells, and benign epithelium. This approach enhances the potential for precise clinical biomarker discovery with high-resolution, whole transcriptome spatial transcriptomics and a deeper understanding of the TME.

A longitudinal, circulating tumor molecular response biomarker as a predictor of clinical outcomes in a real-world advanced pan-cancer cohort of patients treated with tyrosine kinase inhibitors

In a study analyzing advanced cancer patients, researchers evaluated the prognostic value of changes in circulating tumor DNA tumor fraction (ctDNA TF) during tyrosine kinase inhibitor (TKi) therapy. The study, which consisted of 109 patients from Tempus' multimodal real-world database, found that molecular responders had significantly longer real-world overall survival (rwOS) than molecular non-responders across various cancer types. The findings suggest that ctDNA TF may serve as a biomarker to predict molecular response to TKi therapy, potentially guiding treatment decisions and improving patient outcomes in a real-world setting.

A novel approach to define ctDNA molecular response to immunotherapy

Within Tempus' multimodal real-world database, researchers identified 71 pan-cancer patients undergoing immune checkpoint inhibitor (ICI) therapy. The team used circulating tumor DNA tumor fraction (ctDNA TF) as a biomarker to classify molecular responders (MRs) and non-responders (nMRs). Results indicated that MRs had significantly longer real-world overall survival (rwOS) compared to nMRs. Further, the study suggests that patients who maintained a ctDNA TF below a specific low TF threshold, irrespective of the values observed, had prolonged real-world overall survival (rwOS), comparable to MRs. This finding suggests that patients maintaining low TF levels represent a prognostically favorable subgroup, warranting prospective validation of this threshold for early treatment intervention.

Functional precision medicine: Uncovering high actionability in rare cancers beyond genomics

This study assesses the potential of ex-vivo drug sensitivity testing to enhance precision medicine, particularly for patients with rare cancers. Utilizing the PARIS® test, researchers evaluated patient-derived tumor cells (PDTCs) against a customized oncology drug panel based on cancer type, oncologist recommendations, and genomic features. The study involved 86 patients with rare cancers, achieving an 85% actionability rate, indicating at least one drug showed exceptional or good responses. One LGSOC patient achieved a 29-month disease stabilization using PARIS®-informed therapy. The findings demonstrate that the PARIS® test offers unbiased ex-vivo drug sensitivity testing with the potential to identify treatment options, which may be particularly valuable in guiding therapy selection for rare cancers with limited standard treatment avenues.

Identification of predictive biomarkers of response to ADCs by HTS of highly molecularly characterized panels of patient-derived organoids (PDOs)

In a study aimed at enhancing the understanding of the efficacy of antibody-drug conjugates (ADCs), researchers developed a platform to generate and screen a diverse panel of patient-derived organoids (PDOs) across 10 solid tumor indications. Through culturing and viability assays, the team observed ADC-specific responses in the 60 PDO models tested, with target expression being a primary driver of sensitivity. Differential gene expression analysis uncovered additional potential biomarkers, suggesting a complex molecular landscape of ADC response. These findings underscore the value of PDO screening in identifying molecular factors that contribute to ADC sensitivity and resistance.

Molecular subtypes identified in lung adenocarcinoma using a large-scale, real-world dataset

This study explores the molecular diversity of lung adenocarcinoma (LUAD) by analyzing data from Tempus' multimodal real-world database that includes both early and late-stage patients. Researchers at Tempus performed transcriptomic analysis to identify six distinct molecular subtypes (C1-C6) of LUAD, each with unique mutational characteristics, tumor microenvironment (TME) profiles, and real-world progression-free survival (PFS) outcomes. These findings underscore the complexity and heterogeneity of LUAD, offering insights that could potentially guide personalized treatment strategies and inform the development of new therapeutic approaches tailored to specific molecular subtypes.

Multi-modal Large Language Models for Metastatic Breast Cancer Prognosis

This study introduces a novel approach to cancer prognosis by transforming structured data into a narrative format akin to clinical notes. Large language model (LLM) consume the notes to predict overall survival (OS) in metastatic breast cancer (mBC) patients. The Patient Chronological Note (PCN) algorithm, combined with DistilBERT, an LLM pre-trained on extensive breast cancer patient data, resulted in a model that outperformed a standard linear cox model (LLM-Cox). The LLM-Cox model’s interpretable embeddings identified ten distinct patient clusters with varying prognostic risks, offering a new avenue for patient stratification and the opportunity to identify novel treatment-specific biomarkers.

Association between LAG3 expression and immune checkpoint inhibitor (ICI) efficacy in advanced melanoma

Predictive biomarkers for first-line (1L) anti(a)-PD1 alone or in combination with aLAG-3 or aCTLA4 ICIs in melanoma are limited. Furthermore, the relationship between LAG3 expression and clinical benefit to 1L non-aLAG-3 ICIs is unknown. This study investigates the potential of LAG3 expression as a predictive biomarker for first-line (1L) immune checkpoint inhibitors (ICIs) in advanced melanoma, including aCTLA4/PD1 combination therapy versus aPD1 monotherapy. In a real-world cohort of patients with advanced melanoma from the Tempus multimodal database, researchers found that high LAG3 mRNA expression correlated with PD-L1 positivity and increased adaptive immune cells. Importantly, LAG3 levels altered the efficacy of aPD1 agents but not aCTLA4/PD1 combination therapy as measured by real-world overall response rate and overall survival. Prospective studies should validate these novel findings in advanced melanoma to identify whether 1L aCTLA4/PD1 or alternative ICIs are optimal for patients based on LAG3 expression levels.

Comprehensive Whole Genome Sequencing (WGS) Assay Provides Diagnostic Insight into Clinically Relevant Genomic Alterations Across Myeloid Malignancies

The research team evaluated whole-genome sequencing (WGS) as a diagnostic tool for myeloid malignancies. Analyzing 230 patients, the WGS assay identified key genetic variations and demonstrated high concordance with conventional methods in identifying guideline-recommended genomic alterations. The findings suggest that WGS has the potential to streamline the diagnostic process, reduce costs, and improve personalized treatment strategies, especially in regions with limited access to cytogenetic testing.

Low interferon expression coupled with 9p21 loss in non-small cell lung cancer (NSCLC) patients associated with distinct somatic landscape, altered immune population, and poorer response to immune checkpoint inhibitor (ICI) therapies

This study explores the implications of 9p21 deletions and their association with immune phenotypes and treatment outcomes. Tempus' multimodal real-world database was utilized to analyze NSCLC patients with combined CDKN2A/MTAP deletions and varying levels of IFNK expression. The analysis of 16,947 patients revealed that those with CDKN2A/B/MTAP deletions were more likely to exhibit low IFNK expression, which trended towards a higher rate of progressive disease (PD) following immune checkpoint inhibitor (ICI) therapies. The study also identified a relationship between CDKN2A/MTAP deletions and driver alterations in EGFR and ALK.

Stratification of Cell Therapies in Solid Tumor Organoids Using Deep Learning-Derived Imaging Metrics

The team utilized deep learning algorithms to predict the viability of patient-derived organoids (PDOs) from brightfield images, and to extract a set of interpretable phenotypes of response, capturing the impact of 27 different cell therapies across 60 PDOs covering 9 cancer types. The brightfield model was trained on over 11,000 time-lapse images and demonstrated high concordance with ground truth viability measures. This eliminated the need for fluorescent vital dye stains and enabled dynamic response readouts without impairing cell function. Additionally, the interpretable features of response clustered therapies into functional groups, capturing differential mechanistic effects such as infiltration and apoptosis dynamics. This offered a more comprehensive view of therapy efficacy compared to standard terminal viability readouts. This scalable platform offers a promising tool for high-throughput screening of cell therapies, enabling the stratification of treatments into meaningful categories based on their biological effects on PDOs.

Investigating the clinical landscape and biological impact of SF3B1 hotspot mutations in breast cancer

This study examines the implications of SF3B1 hotspot mutations in breast cancer, focusing on their genetic profile, survival outcomes, and biological impacts, by analyzing de-identified data from Tempus’ multimodal real-world database consisting of 420 breast cancer patients with SF3B1 mutations. Innovative genome editing in isogenic breast cell lines revealed that SF3B1 mutations negatively impact cell growth and tumor development. The findings support the utility of SF3B1 mutations as potential therapeutic targets and underscore the importance of understanding their role in cancer biology, with ongoing research aimed at uncovering the mechanisms behind hotspot-specific effects.