Abstract Effective clinical management of cancer patients requires an accurate and early diagnosis, highly sensitive monitoring of minimal residual disease (MRD), and precise therapy selection. There are multiple tests available that attempt to address each of these needs independently with varying degrees of clinical utility. Caris Assure is a proprietary circulating nucleic acid sequencing platform that couples whole exome and transcriptome (WES/WTS) sequencing on white blood cells and plasma with advanced machine learning techniques to satisfy all three testing needs on one platform. This test detects SNVs, INDELs, structural variants, copy number, gene expression, tumor mutational burden (TMB), microsatellite instability (MSI), fragment length, and aneuploidy of both somatic (tumor and clonal hematopoiesis) and germline origin. Caris’ extensive database of over 350,000 tissue WES/WTS from solid malignancies was used to train deep learning neural networks to identify the molecular underpinnings of cancer. These networks were then deployed on WES/WTS data from plasma and buffy coat in pursuit of signals that can inform early detection, MRD and therapy selection. Validation studies were performed to characterize the analytic and clinical performance of Assure on over 3000 patient blood samples. These samples include ~1000 non-cancer patients (controls), ~1700 newly diagnosed patients where blood was collected at surgery (early detection), ~500 early-stage patients during adjuvant therapy at multiple time points (MRD), and ~200 locally advanced/metastatic patients where matched tissue testing was also performed (therapy selection). For early detection, stratification of blood samples from patients with stage I-IV cancer versus those with no reported cancer resulted in an AUC > 0.99 and included over 30 types of solid tumors. Notably, at 99.5% specificity, the sensitivities for stages I-IV (n= 119, 54, 50, 27) were 73%, 80%, 76%, and 89%. In the MRD setting for high-risk patients, the disease-free survival of patients whose cancers were predicted to recur was significantly shorter (39.6 mo) than those predicted not to recur (93.4 mo) (HR: 5.18, 95%CI: 2.94-9.09, p<.00001). This performance was observed across multiple lineages of cancer including but not limited to breast, colon, lung, and bladder. Lastly, for therapy selection, detection of driver mutations where blood was collected within 30 days of matched tissue demonstrated high concordance with a PPA of 93.8% and PPV of 96.8%. CHIP correction proved to be essential as ~35% percent of patients had CHIP mutations, including KRAS, BRAF, ATM, BRCA1/2, findings that could lead to improper therapy selection. Herein, we demonstrate for the first time a single liquid biopsy assay that addresses the entire continuum of care in clinical oncology with optimal diagnostic, prognostic, and predictive utility for patients and physicians. Citation Format: Jim Abraham, Valeriy Domenyuk, Maria Perdigones Borderias, Takayuki Yoshino, Elisabeth I. Heath, Emil Lou, Stephen Liu, John Marshall, Wafik S. El-Deiry, Anthony Shields, Martin Dietrich, Yoshiaki Nakamura, Takao Fujisawa, David D. Halbert, Dominic Sacchetti, Seth Stahl, Adam Stark, Sergey Klimov, Sourabh Antani, Chadi Nabhan, Jeffrey Swensen, George Poste, Matt Oberley, Milan Radovich, George W. Sledge, David Spetzler. AI enabled whole exome& transcriptome liquid biopsy addressing MCED, MRD, and therapy selection on a single platform [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2024; Part 1 (Regular Abstracts); 2024 Apr 5-10; San Diego, CA. Philadelphia (PA): AACR; Cancer Res 2024;84(6_Suppl):Abstract nr 2300.

Demographics of TRIBE2 RCT cases used in this study compared with the full trial.

Kaplan-Meier plots of probability of overall survival in patients with advanced PDAC receiving first-line FOLFIRINOX (A) or nab-paclitaxel/gemcitabine (B) or patients with GC/EGJC receiving an oxaliplatin-containing regimen (C). A) The median OS in the Increased Benefit cohort is 10.1 months (90%) longer than the median OS in the Decreased Benefit cohort (HR = 0.478, 95% CI: 0.289-0.792, log-rank p = 0.003). B) The median OS in the Increased Benefit cohort is one month (11%) longer than the median OS in the Decreased Benefit cohort (HR = 0.958, 95% CI: 0.658-1.395, log-rank p = 0.823). C) The median OS in the Increased Benefit cohort is 5.1 months (58%) longer than the median OS in the Decreased Benefit cohort (HR = 0.437, 95% CI: 0.250-0.763, log-rank p = 0.003).

A) Correlation of TTNT and PFS of cases in the TRIBE2 trial. B) Distribution of mCRC RWE TNTs compared with PFS from the FOLFOX and FOLFOXIRI arms of the TRIBE2 trial.

Demographics of the cases used in the PDAC RWE testing set.

Cross-validated performance of the training cases. The median TTNT in the Increased Benefit cohort is 3.9 months (52%) longer than the median TTNT in the Decreased Benefit arm (HR = 0.398, 95% CI: 0.244-0.647, log-rank p < 0.001).

Demographics of the cases used in algorithm training.

Demographics of the cases used in the EC/GEJC RWE testing set.

e15049 Background: Liquid biopsies have become integral to the management of patients with advanced malignancies, however significant deficiencies exist in currently used platforms. Liquid biopsy data often show unacceptably high discordance compared to gold standard tissue-based data; this is often due to false positives (FP) resulting from clonal hematopoiesis (CH) or increased false negatives due to inherently poor sensitivity as well as the confounding intra-patient or tumor stage dependent ‘non-shedder’ problem. Caris Assure is a novel whole exome (WES) and whole transcriptome (WTS) NGS assay performed on plasma and buffy coat (BC) fractions of whole blood that maximizes the number of unique nucleic acids, thereby increasing sensitivity as well as identifying incidental germline variants and correcting for CH. Methods: Matched tissue and whole blood specimens were obtained from 184 patients with solid tumors. Whole blood was collected in PAXgene ccfDNA tubes. DNA & RNA were co-extracted from matched plasma, BC, and FFPE tumor tissue specimens and prepared for NGS using hybrid bait capture methodology. Plasma and BC samples were tested using the Caris Assure WES/WTS assay as well as orthogonal assays. Tissue specimens were tested using Caris’ validated WES/WTS tissue assay. Results were compared between the plasma assays and matched tissues as well as using a majority-rules calling model. A positive plasma assay variant was counted a true positive if it was also identified in the matched tissue or the orthogonal plasma result. Similarly, a negative result was considered a true negative if either of the other two test results were also negative. Results: When data were analyzed from patients with metastatic disease and were focused on only those genes included in the limited orthogonal liquid panel, and tissue was treated as the gold standard, Caris Assure achieved a PPA for SNVs and INDELS of 81.4% and PPV of 85.1%, while the orthogonal assay achieved a PPA and PPV of 79.1% and 53.1%. When the majority rules approach was used, Assure PPA and PPV were 93.3% and 93.3%, respectively, compared to 91.1% and 64.1%. Germline variants detected by Caris Assure had 100% PPA & PPV when compared to the external germline assay. There were 444 pathogenic/likely pathogenic mutations detected across all assays. Of those, 77 (17.3%) were derived from CH and identified as such in the Caris Assure results, though they are inaccurately reported as tumor-derived variants using the comparator assay. Using Caris Assure corrected for CH-derived false positives, the correlation between plasma and tissue TMB was very strong (r = 0.9) when tissue and blood specimens were collected within 4 weeks of each other. Conclusions: The novel Caris Assure liquid biopsy platform demonstrates high sensitivity and specificity for somatic variants, while mitigating against FPs from CH, and identifying incidental germline pathogenic alterations.

AbstractPurpose: FOLFOX, FOLFIRI, or FOLFOXIRI chemotherapy with bevacizumab is considered standard first-line treatment option for patients with metastatic colorectal cancer (mCRC). We developed and validated a molecular signature predictive of efficacy of oxaliplatin-based chemotherapy combined with bevacizumab in patients with mCRC. Experimental Design: A machine-learning approach was applied and tested on clinical and next-generation sequencing data from a real-world evidence (RWE) dataset and samples from the prospective TRIBE2 study resulting in identification of a molecular signature, FOLFOXai. Algorithm training considered time-to-next treatment (TTNT). Validation studies used TTNT, progression-free survival, and overall survival (OS) as the primary endpoints. Results: A 67-gene signature was cross-validated in a training cohort (N = 105) which demonstrated the ability of FOLFOXai to distinguish FOLFOX-treated patients with mCRC with increased benefit from those with decreased benefit. The signature was predictive of TTNT and OS in an independent RWE dataset of 412 patients who had received FOLFOX/bevacizumab in first line and inversely predictive of survival in RWE data from 55 patients who had received first-line FOLFIRI. Blinded analysis of TRIBE2 samples confirmed that FOLFOXai was predictive of OS in both oxaliplatin-containing arms (FOLFOX HR, 0.629; P = 0.04 and FOLFOXIRI HR, 0.483; P = 0.02). FOLFOXai was also predictive of treatment benefit from oxaliplatin-containing regimens in advanced esophageal/gastro-esophageal junction cancers, as well as pancreatic ductal adenocarcinoma. Conclusions: Application of FOLFOXai could lead to improvements of treatment outcomes for patients with mCRC and other cancers because patients predicted to have less benefit from oxaliplatin-containing regimens might benefit from alternative regimens.

Spliceosomal dysregulation dramatically affects many cellular processes, notably signal transduction, metabolism, and proliferation, and has led to the concept of targeting intracellular spliceosomal proteins to combat cancer. Here we show that a subset of lymphoma cells displays a spliceosomal complex on their surface, which we term surface spliceosomal complex (SSC). The SSC consists of at least 13 core components and was discovered as the binding target of the non-Hodgkin's lymphoma-specific aptamer C10.36. The aptamer triggers SSC internalization, causing global changes in alternative splicing patterns that eventually lead to necrotic cell death. Our study reveals an exceptional spatial arrangement of a spliceosomal complex and defines it not only as a potential target of anti-cancer drugs, but also suggests that its localization plays a fundamental role in cell survival.

Abstract Introduction: Deconvolution of multi-nodal perturbations in cancer network architecture demands highly multiplexed profiling assays. We demonstrate the value of polyligand profiling of tumor systems states using libraries of single stranded oligodeoxynucleotides (ssODN) to distinguish between tumor tissue from breast cancer patients who did or did not derive benefit from treatment regimens containing trastuzumab. Methods: This study included cases from women with invasive breast cancer who received chemotherapy+ trastuzumab (C+T) or trastuzumab monotherapy with available retrospective data on the time to next treatment (TTNT). A library of 2x1012 unique ssODN was exposed to FFPE tissues from patients who benefited (B) or not (NB) from trastuzumab-based regimens in several rounds of positive and negative selection. Two enriched libraries were screened on independent set of 42 B and 19 NB cases using a modified IHC protocol for detection of bound ssODNs. Poly-Ligand Profiles (PLP) were scored by a blinded pathologist. Two libraries, EL-NB and EL-B, showed significant p-values between groups of responders and non-responders. A Cox-PH model was fitted using either tumors' HER2 status or PLP test results as the independent variable. Median survival time was calculated from the Kaplan-Meier estimate. A separate group of 63 cases with TTNT data from chemotherapy without trastuzumab was used as a control to distinguish prognostic from predictive performance. Results: The PLP scores of EL-NB and EL-B were assessed by receiver operating characteristic (ROC) curves and resulted in a combined AUC value of 0.81. EL-NB and EL-B were able to effectively classify B and NB patients with either HER2-negative/equivocal (AUC = 0.73) or HER2-positive cancers (AUC = 0.84). In contrast, HER2 status alone yielded an AUC value of 0.47. The combined PLP scores for the independent set of 63 patients treated with C excluding trastuzumab resulted in an AUC value of 0.53, indicating that the assay was predictive and not simply prognostic. Kaplan-Meier curves analysis shows that PLP+ cases have 429 days median TTNT, while PLP- cases have 129 days (HR = 0.38, log-rank p = 0.001). Analysis based on HER2 status showed no significant difference in TTNT between patients that were HER2+ (280 days) or HER2-negative/equivocal (336 days, HR = 1.27, log-rank p =0.45). Summary: Performance of the PLP assay in differentiating patients who did or did not benefit from trastuzumab therapy outperforms the standard IHC assay for HER2 status. These results represent a promising step towards the development of a CDx to identify the 50-70% of HER2+ patients who will not benefit from trastuzumab. In addition, PLP also has the potential to identify the HER2-negative/equivocal patients who may benefit from trastuzumab-containing regimens. Citation Format: Domenyuk V, Gatalica Z, Santhanam R, Wei X, Stark A, Kennedy P, Toussaint B, Levenberg S, Wang R, Xiao N, Greil R, Rinnerthaler G, Gampenrieder S, Heimberger AB, Berry DJ, Barker A, Demetri GD, Quackenbush J, Marshall JL, Poste G, Vacirca JL, Vidal GA, Schwartzberg LS, Halbert DD, Voss A, Miglarese MR, Famulok M, Mayer G, Spetzler D. Polyligand profiling differentiates cancer patients according to their benefit of treatment [abstract]. In: Proceedings of the 2017 San Antonio Breast Cancer Symposium; 2017 Dec 5-9; San Antonio, TX. Philadelphia (PA): AACR; Cancer Res 2018;78(4 Suppl):Abstract nr P2-09-09.

Assessing the phenotypic diversity underlying tumour progression requires the identification of variations in the respective molecular interaction networks. Here we report proof-of-concept for a platform called poly-ligand profiling (PLP) that surveys these system states and distinguishes breast cancer patients who did or did not derive benefit from trastuzumab. We perform tissue-SELEX on breast cancer specimens to enrich single-stranded DNA (ssDNA) libraries that preferentially interact with molecular components associated with the two clinical phenotypes. Testing of independent sample sets verifies the ability of PLP to classify trastuzumab-treated patients according to their clinical outcomes with ROC-AUC of 0.78. Standard HER2 testing of the same patients gives a ROC-AUC of 0.47. Kaplan-Meier analysis reveals a median increase in benefit from trastuzumab-containing treatments of 300 days for PLP-positive compared to PLP-negative patients. If prospectively validated, PLP may increase success rates in precision oncology and clinical trials, thus improving both patient care and drug development.

12067 Background: The MAESTRO trial randomized 693 locally advanced or metastatic pancreatic cancer patients to gemcitabine (G) + placebo vs G + evofosfamide (GE). OS hazard ratio (HR) was 0.84; p = 0.059. We developed a PLP assay that identifies patients most likely to benefit from GE vs G alone. By capitalizing on ssDNA aptamer binding properties, PLP measures network changes in tumors, including those that predict drug response. Methods: FFPE tissues of pancreatic cancer patients from the MAESTRO trial with good (OS > 13 mos) or poor (OS < 7 mos) outcome from GE were used for PLP assay development. Assay cut-points were determined using a training set (n = 12) and performance metrics were then determined using an independent blinded test set (n = 172). The study population enrolled in MAESTRO was divided into four cohorts based on treatment and benefit (cut-point = 240 days). The assay performance from the blinded test set was used to generate 1000 different possible patient subsets. Each simulation represents one possible trial outcome if the PLP assay had been used to enroll patients. We used the average value of the simulated median increase in OS from the 1000 random selections to estimate the impact the PLP assay would have had on MAESTRO. Results: 97% of the simulations yielded log-rank p < 0.05. Compared to MAESTRO, the average median OS increase for GE improved by 116±38% (simulation s.d.) with an average HR of 0.72±0.04. Moreover, the sample size in the simulations was 49% smaller compared to the original study, reflecting the percentage of test positive patients in the blinded test set. When only data from primary tumors were used, 100% of the permutations yielded log-rank p < 0.05 with an average median OS increase of 216±36% compared to MAESTRO and an average HR of 0.63±0.03. Conclusions: This retrospective study demonstrates that PLP, if prospectively applied, likely would have resulted in a successful study comparing GE to G pancreatic cancer patients. PLP is a powerful, flexible and facile platform warrants further study of additional therapies and in prospective trials.

Technologies capable of characterizing the full breadth of cellular systems need to be able to measure millions of proteins, isoforms, and complexes simultaneously. We describe an approach that fulfils this criterion: Adaptive Dynamic Artificial Poly-ligand Targeting (ADAPT). ADAPT employs an enriched library of single-stranded oligodeoxynucleotides (ssODNs) to profile complex biological samples, thus achieving an unprecedented coverage of system-wide, native biomolecules. We used ADAPT as a highly specific profiling tool that distinguishes women with or without breast cancer based on circulating exosomes in their blood. To develop ADAPT, we enriched a library of ~10 11 ssODNs for those associating with exosomes from breast cancer patients or controls. The resulting 10 6 enriched ssODNs were then profiled against plasma from independent groups of healthy and breast cancer-positive women. ssODN-mediated affinity purification and mass spectrometry identified low-abundance exosome-associated proteins and protein complexes, some with known significance in both normal homeostasis and disease. Sequencing of the recovered ssODNs provided quantitative measures that were used to build highly accurate multi-analyte signatures for patient classification. Probing plasma from 500 subjects with a smaller subset of 2000 resynthesized ssODNs stratified healthy, breast biopsy-negative, and -positive women. An AUC of 0.73 was obtained when comparing healthy donors with biopsy-positive patients.

e23070 Background: Tissue biopsies, required for definitive diagnosis of cancer, are risky and costly. We developed the non-invasive screening method ADAPT (Adaptive Dynamic Artificial Poly-ligand Targeting). Here we report the feasibility of aptamer library enrichment on blood plasma, which is primarily driven by exosome-associated proteins that may serve as biomarkers and drug targets. Methods: ssDNA libraries of ~10^12 oligodeoxynucleotides (ODNs) were enriched for associating with exosomes present in plasma. Specific interactions of enriched libraries and exosomes were evaluated by flow cytometry, NGS, qPCR and LC-MS/MS. Results: To validate the enrichment process is capable of identifying exosome specific ODNs in plasma, we titrated cancer cell line exosomes into human plasma and performed independent enrichments. qPCR and NGS revealed a linear relationship between exosomal input and the number of recovered unique ODNs. We then applied this enrichment scheme to human clinical samples of breast cancer/normal plasma. Flow cytometry showed that the enriched library bound preferentially to human plasma exosomes. Using NGS, this result was confirmed by comparing the binding profile of the enriched library recovered from unfractionated versus exosome depleted plasma from individual patients. Library mediated affinity purification of plasma exosomes and subsequent MS analysis revealed 96 exosome-associated proteins that were not detectable otherwise. Compared to the unenriched library, the breast cancer plasma enriched library showed higher binding to MCF7 breast cancer cells than to HS578Bst normal breast epithelial cells. This result shows that plasma-based enrichment trained the library towards recognizing proteins associated with the malignant phenotype. Conclusions: We have demonstrated that ADAPT applied to human plasma results in the enrichment and identification of ODNs with preferential binding to cancer exosomes. The detection of low-abundance proteins in human plasma suggests that ADAPT may be leveraged for biomarker and drug target identification.

e23058 Background: Understanding individual biomarkers and multi-molecular complexes in their native states represents a major hurdle in the development of systems biology platforms and requires multiplexing capabilities many orders of magnitude greater than what is currently available. This knowledge is especially important in diseases like cancer, where perturbations in signaling pathways lead to the initiation and propagation of a vast range of molecularly heterogeneous phenotypes. Ideally, this information would be gathered non-invasively from peripheral compartments. We report ADAPT as an unbiased discovery method to identify native proteins and multi-molecular complexes in plasma. Methods: ssDNA libraries of 2x10^11 unique sequences were enriched towards association with exosomes of plasma pools from breast cancer patients (n=60) or from a control cohort of women without breast cancer (n=60). Two thousand ODNs were selected from the enriched library and used to profile 500 plasma samples: 206 breast cancer biopsy positive (BC+), 177 biopsy negative (BC-) and 117 self-declared healthy (H). Results: Random Forest Modeling (RFM) was used to build classifiers for BC+ vs. both BC- and H, BC+ vs. BC- and BC+ vs. H, with AUC values of 0.64, 0.64, and 0.73, respectively. Out-of-bag validation was used for this analysis. Notably, 10-fold cross-validation, repeated 20 times, resulted in ROC AUC values of 0.58 for BC+ vs. both controls, 0.59 for BC+ vs. BC-, 0.62 for BC+ vs. H. This difference between the out-of-bag and cross-validation result suggests that the study is underpowered to define an algorithm capable of handling the heterogeneity found in this set of breast cancer patients. Conclusions: The ADAPT platform uses ultra-high complexity libraries to improve the probability of identifying ODNs which recognize biomolecules in their native state. This unbiased approach quantifies differences between plasma from women with breast cancer and women without breast cancer, likely due to perturbations in molecular pathways. An ADAPT derived breast cancer test may find utility as a diagnostic tool in clinical practice.

e22052 Background: Molecular profiling-guided treatment strategies are frequently being used when other options do not exist. The availability of data generated by next-generation sequencing (NGS) has ushered in a new era of molecular profiling and needs to be evaluated for use in making clinical treatment decisions. As such, we propose a definition for actionable molecular data and show the frequency of actionable alterations across a variety of platforms including immunohistochemistry (IHC), in situhybridization (ISH), and NGS. Methods: We evaluated 12,265 patients using 34 actionable biomarkers to make associations with 43 FDA-approved drugs. We define a biomarker alteration as actionable when the association between the alteration and the drug is supported by clinical evidence. In this analysis, clinical evidence was demonstrated by a randomized prospective study, a nonrandomized cohort/case-controlled retrospective or an observational study with statistically significant associations showing predictive utility. We did not deem biomarkers and drugs connected only by mechanistic associations (e.g., a common cellular pathway) as actionable, as treating patients with compounds not demonstrated to be effective by clinical evidence may not be optimal. Results: Using the above definition of actionable, we found that recommendations could be made for drugs with a potential beneficial response for 94% of patients and a recommendation against drugs with a potential lack of benefit could be made for 97% of patients. Of the beneficial compounds identified, 87% were driven by IHC and ISH results, 13% by IHC, ISH, and NGS results, and 0.2% by NGS results alone. Of the lack of benefit compounds identified, 74% were driven by IHC and ISH, 20% by IHC, ISH, and NGS, and 6% by NGS alone. Conclusions: These results show that a multiplatform approach to molecular profiling is essential to provide patients with the clinically supported information needed to help treat their disease with broadly available therapies. Using NGS alone misses the majority of actionable alterations, but may prove more beneficial when the next generation of targeted therapies is available.