Tang, Xiaojia

Abstract Background: The increasing volume and diversity of clinical genomic reports pose a significant challenge across healthcare institutions for the interpretation and proper implementation of precision oncology research. Reports from multiple different vendors (e.g., Invitae, Ambry Genetics, Foundation Medicine) are typically PDFs and exhibit substantial heterogeneity in panel design, gene coverage, and reporting standards, which hinders efficient retrospective data mining and patient cohort identification within the electronic medical records. We sought to develop a framework to interrogate all of these reports. Methods: We extracted genomic reports from patients with breast cancer treated with neoadjuvant chemotherapy and developed MolHarmonizer, a novel, scalable framework leveraging Python and Gemini LLMs, designed to process and harmonize genomic data from disparate multi-vendor reports. Gemini LLMs are employed explicitly for robust information extraction, normalization, and structuring of key genomic features, transforming unstructured data into a unified, queryable dataset. Results: Our MolHarmonizer framework successfully processed 1,147 genomic reports from 1703 breast cancer patients (2006-2023) from 23 different companies, demonstrating robust capability to extract and standardize critical actionable biomarkers. Data sources included Invitae (n=554), Ambry Genetics (n=189), Natera (n=95), Mayo Clinic (n=88), Tempus (n=63), Guardant Health (n=47), and Foundation Medicine (n=37), with others contributing less than 20 reports. Of the samples, 827/1147 (72.1%) were germline (blood/saliva). A majority of the patients were tested using the panels due to a personal/family history (n=925). Overall, 413/1147 (36.0%) reports identified at least one mutation. For breast cancer, 75 reports showed BRCA1/2 mutations (37 BRCA1, 37 BRCA2, and one patient with both BRCA1 and 2). Other mutations identified included: PIK3CA (n=40), TP53 (n=96), PTEN (n=21), ESR1 (n=13) and AKT1/2 (n=8). Conclusion: MolHarmonizer, a powerful framework leveraging Gemini LLMs, effectively addresses genomic data heterogeneity by automating biomarker extraction and harmonization. This enables rapid cohort identification and deep retrospective analyses for clinical insights, biomarker discovery, understanding disease history, facilitating novel pattern discovery, e.g., predicting BRCA1 mutations from WSI, and accelerating research within our neoadjuvant BC cohort. Future plans include expanding to include over 20,000 breast cancer patients, developing a user-friendly chatbot, and ensuring inter-institutional adaptability for a variety of complex diseases. Citation Format: Krishna Rani Kalari, Xiaojia Tang, Thanmayee Boyapati, Tanya L. Hoskin, Sumathilatha Myla, Sumedha G. Penheiter, Richard M. Weinshilboum, Liewei Wang, Hamid R. Tizhoosh, Karthik Vikram Giridhar, Matthew P. Goetz, Judy C. Boughey. Development of an LLM framework for analysis of heterogeneous breast cancer patients genomic reports [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2026; Part 1 (Regular Abstracts); 2026 Apr 17-22; San Diego, CA. Philadelphia (PA): AACR; Cancer Res 2026;86(7 Suppl):Abstract nr 2756.

e14545 Background: Next generation sequencing (NGS) identifies genomic alterations in DNA obtained from tissue biopsies and/or blood [circulating tumor (ct)DNA]. We anticipate a lower rate of copy number loss (CNL) in ctDNA compared to tissue, as lower ctDNA tumor fraction may limit the ability to detect a deletion. However, we observed several instances of a CNL in ctDNA that was not identified in a tissue specimen, prompting this systematic evaluation of an institutional dataset. We focused on BRCA1/2 pathogenic variants (PVs) given their therapeutic implications and coverage across testing platforms. Methods: Using an IRB-approved protocol, we evaluated an institutional repository of NGS results performed across the Mayo Clinic Enterprise. We focused on patients (pts) who had both tissue and blood NGS testing performed. We further narrowed this cohort to those with a BRCA1/2 PVs, somatic and/or germline, as determined by testing performed at Guardant Health, Tempus, or Foundation Medicine. We abstracted information about each variant, including type of alteration classification and functional impact. Demographic information, cancer characteristics, and germline testing information was collected. Fisher’s exact test was used to compare alteration frequencies in ctDNA vs tissue using GraphPad Prism v 10.1. Results: From Dec 2015-Sept 2024, we identified 158 unique pts with BRCA1/2 PV who had NGS results for both tissue and ctDNA. The median age at first reporting BRCA1/2 variant was 67 years (range: 29-98 years). 48% of pts were female, and the most common cancer type sites were lung (23%), colorectal (20%), and pancreaticobiliary (20%). We identified 198 unique pathogenic genomic alterations; 71 in BRCA1 & 127 in BRCA2 . Most common alteration was an SNV or indel (127/198), followed by CNL (n = 63), splice site variants (n = 5), fusions (n = 2), and rearrangements (n = 1). Of the 126 SNV or indels, 50 (39.7%) were identified in both ctDNA and tissue, 40 (31.7%) were detected in blood only, 36 (28.6%) identified only in tissue (p < 0.0001). Amongst the 63 CNL, only 3 (4.8%) were identified in both ctDNA and tissue; 41(65.1%) were only identified in ctDNA, and 19 (30.2%) in tissue only (p < 0.0001). In the 19 tissue-only CNL tests, 12/19 noted one or more additional pathogenic SNV/Indel in BRCA1/2. In the 41 ctDNA-only CNL tests, 1/41 noted one or more additional pathogenic SNV/Indel in BRCA 1/2 . Germline testing was available in 41/63 pts with copy number loss. 3/41 deletions were identified in germline (2 in ctDNA, 1 in tissue). Conclusions: We observed a low degree of concordance in CNL between tissue and ctDNA for BRCA1/2 . CNL detection in ctDNA that could not be confirmed in germline or with a second alteration warrants further attention. In particular, clinicians should be cautioned when interpreting ctDNA reports noting BRCA1/2 CNL without clarity if this represents biallelic loss or a synchronous loss of function.

Abstract Deep learning has advanced digital pathology and multi-omics data analysis. However, there is still a critical need to integrate these modalities to better understand complex diseases. Deep learning models, such as Prov-Gigapath, a state-of-the-art pathology foundation model, have been developed for analyzing whole-slide images (WSI). Our previously published deep-learning framework, OmicsFootPrint (OFP), transforms high-dimensional multi-omics data into intuitive genomically-arranged circular images and utilizes transfer learning to overcome the challenges of high-dimensional, low-sample-size data. In this study, we aimed to combine WSI and multi-omics data to develop a framework capable of distinguishing between invasive ductal carcinoma (IDC) and invasive lobular carcinoma (ILC), which is clinically significant as it may help explain differential responses to therapies, such as tamoxifen versus letrozole. We enhanced the performance of our OFP model by dividing each multi-omics circular image (1024x1024x3 pixels) into 16 equal-sized patches (256x256x3 pixels). Each patch was trained using a separate EfficientNetV2 architecture with pre-trained ImageNet weights. The probability outcomes from these 16 models were subsequently combined through a meta-learner model using the AutoGluon framework. Concurrently, WSI images corresponding to the same patients were independently trained, maintaining consistent train/validation/test splits. The Prov-Gigapath model, which was trained using a dataset from the Providence Health System, was employed to extract patch-level embeddings from the WSI data. These embeddings were fine-tuned using a LongNet-based masked autoencoder to predict outcomes. Finally, predictions from the OFP model and the WSI-based Prov-Gigapath model were integrated through a meta-learner, preserving the original train/validation/test split. Model performances (AUC) were evaluated using the held-out test set. The gold standard used for the analysis was histological identification of IDC versus ILC, against which the performance of omics data alone, WSI data alone, and the multi-modal integration approach were compared. We evaluated our integrated model as a proof-of-concept using The Cancer Genome Atlas (TCGA) breast cancer cohort, comprising 92 ILC and 391 IDC samples. Multi-omics data, including gene expression, microRNA, copy number variation, and RPPA, were analyzed using the OFP framework, while WSI data were processed using the Prov-Gigapath method. The results demonstrated that the multi-modal integration approach significantly outperformed individual modalities, achieving an AUC of 0.98 compared to 0.87 for OFP (original), 0.95 for OFP (16-patches), and 0.92 for WSI. Integrating WSI and multi-omics data improves predictive accuracy, significantly improving the classification of breast cancer subtypes over histology alone. Citation Format: Naresh Prodduturi, Xiaojia Tang, Hamid R. Tizhoosh, Kevin J. Thompson, Richard Weinshilboum, Karthik V. Giridhar, Judy C. Boughey, Eric W. Klee, Liewei Wang, Matthew P. Goetz, Vera Suman, Krishna R. Kalari. A multi-modal learning framework to integrate digital pathology image and multi-omics data in breast cancer [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2025; Part 1 (Regular Abstracts); 2025 Apr 25-30; Chicago, IL. Philadelphia (PA): AACR; Cancer Res 2025;85(8_Suppl_1):Abstract nr 2429.

Abstract Alzheimer’s disease (AD) affects 40 million individuals globally and is characterized by the accumulation of amyloid-beta (Aβ) proteins, which aggregate and form plaques. BBB dysfunction drives AD cerebrovascular pathology and BBB integrity is maintained by neurovascular unit (NVU). Specifically, within the NVU, the cerebral endothelial cells maintain vascular homeostasis. In this study, we isolated endothelial-enriched regions of interest (ROIs) using the Nanostring GeoMx digital spatial profiler and employed a deconvolution model to evaluate transcriptomic changes. We observed dysregulation of cellular signaling potentially disrupting the APP+ BBB integrity. Analysis of ligand-receptor pairings that are the foundation of the NVU intercellular signaling indicated that the endothelial vasculature completes a feedback loop with the NVU to the regulating astroctyes. Further, we identified potentially antagonistic signaling roles for opioid receptor species that should be further investigated for potential therapeutic targets. Graphical Abstract

Abstract The OmicsFootPrint framework addresses the need for advanced multi-omics data analysis methodologies by transforming data into intuitive two-dimensional circular images and facilitating the interpretation of complex diseases. Utilizing deep neural networks and incorporating the SHapley Additive exPlanations algorithm, the framework enhances model interpretability. Tested with The Cancer Genome Atlas data, OmicsFootPrint effectively classified lung and breast cancer subtypes, achieving high area under the curve (AUC) scores—0.98 ± 0.02 for lung cancer subtype differentiation and 0.83 ± 0.07 for breast cancer PAM50 subtypes, and successfully distinguished between invasive lobular and ductal carcinomas in breast cancer, showcasing its robustness. It also demonstrated notable performance in predicting drug responses in cancer cell lines, with a median AUC of 0.74, surpassing nine existing methods. Furthermore, its effectiveness persists even with reduced training sample sizes. OmicsFootPrint marks an enhancement in multi-omics research, offering a novel, efficient and interpretable approach that contributes to a deeper understanding of disease mechanisms.

e13036 Background: In January 2023, the FDA approved elacestrant for ESR1-mutated, ER+/HER2- MBC based on a companion diagnostic cell free (cf)DNA test (Guardant Health). In this study, we analyze the trends of liquid biopsy utilization at Mayo Clinic (MC) over a 3 year period to understand the impact of this drug approval on testing patterns and identify emergent alterations in patients (pts) with MBC. Methods: We extracted clinically performed Guardant 360 and Guardant 360 CDx ordered at MC sites in Rochester (MN), Florida (FL), Arizona (AZ), and across 9 additional Mayo Clinic Health System (MCHS) sites. Pts with MBC were identified using the EMR and analyzing visit diagnosis codes and staging information. Shifts in testing patterns were analyzed over a 3 year period from January 1, 2021 - December 31, 2023. Data was collected on the number of tests, serial testing, frequency of pathogenic mutations, and emergent mutations in ESR1 and PIK3CA. Chi-square test was used to compare the categorical values. Results: Between 2021-2023, 2,741 pts with MBC were seen across all sites. A total of 408 cfDNA tests were performed in 370 unique patients with MBC. Most pts (90%) had 1 test performed. The proportion of pts with MBC tested was 85/1350 (6.3%) in 2021, 77/1495 (5.2%) in 2022, and 208/1558 (13.4%) in 2023. The increased proportion of pts tested in 2023 was statistically significant compared to prior years (p<0.0001), while no difference was observed between 2021 and 2022 (p 0.19). Testing patterns varied by site, with FL ordering the highest number of tests, followed by MN and AZ. Of the 34 providers practicing during the entire period seeing more than 10 MBC pts, the number of providers ordering Guardant testing increased from 15/34 (2021) and 14/34 (2022) to 20/34 (2023), though not statistically significant (p 0.15). Of the 41% with EMR-documented staging available, testing was highest in the ER+/HER2- (81.3%) subtype, followed by TNBC (13.6%) and then HER2+ (5.2%). The most commonly identified mutations identified in the first testing were PIK3CA (41.1%), TP53 (40.5%), ESR1 (30.5%), and ATM (13.2%). D538G (8.0%), Y537N (4.4%), and Y537S (3.5%) were frequent emergent mutations seen in ESR1. Conclusions: We observed an increase in cfDNA testing at the start of 2023 which aligns with the FDA approval of elacestrant. We observed variation by site and individual providers that warrants further exploration to improve cfDNA testing in MBC pts at Mayo Clinic.

1037 Background: TBCRC041, a randomized phase 2 trial of alisertib (A) ± fulvestrant (F) in women with HR+/HER2- MBC demonstrated promising clinical activity for alisertib (1). Interrogation of baseline (PreC1) and end of cycle 1 (EOC1) cfDNA and CTCs may identify biomarkers associated with response to A±F. Methods: Plasma cfDNA was sequenced using the Guardant INFINITY platform, which includes genomic and methylated tumor fraction (mTF) analysis. Pathogenic variants were determined using publicly available databases and Mayo Clinic annotation pipelines. CTCs were identified as nucleated, EpCAM+/ cytokeratin+/ CD45- cells and assessed for ER/HER2 staining (RareCyte, Seattle). PFS was defined as the time from randomization to progression or death. EOC1-PFS differs from PFS as time starts at C2D1. Stratified log-rank tests and conditional Cox modeling were used to assess whether PFS or EOC1-PFS differed by mutations in ESR1 or PIK3CA, pre-CTC ≥ 5, or mTF>1%. Results: Of the 91 eligible pts, 84 progressed, and 73 were deceased. In the 86 pts with PreC1-cfDNA results, variants were identified in ESR1 (42; 48.8%), PIK3CA (36; 41.9%), AKT1 (7; 8.1%), PTEN(7; 8.1%), BRCA1 (4; 4.7%), and BRCA2 (7; 8.1%). Pts with a PIK3CAmut had significantly decreased PFS (HR 1.8; 95%CI: 1.1 -2.9, p=0.014) but not by ESR1mut status (p=0.583). PreC1 and EOC1-mTF were determined for 78 and 76 pts, respectively. PreC1-mTF was 0-1% in 21; 1.1-10.0% in 27; and 10.1-100% in 30. mTF levels remained between 0-1% in 17 pts fell to 0-1% in 11 pts, and was 0-1% in 5 pts without a preC1-mTF determination. PFS did not differ with respect to PreC1-mTF (p=0.210), but EOC1-PFS was worse in pts with mTF >1% (HR 3.0; 95%CI: 1.6 -5.2, p<0.001). PreC1 and EOC1-CTC were determined for 78 and 76 pts. 36 (46.2%) pts had a PreC1 CTC count ≥ 5. EOC1-CTC levels remained at (n=17) or increased to (n=1) ≥ 5 and was ≥ 5 in 2 pts without a preC1-CTC determination. EOC1-PFS was not significantly different for EOC1-CTC ≥ 5 (p=0.065), but was significantly decreased in pts with PreC1-CTC≥5 (HR=1.8; 95%CI: 1.1-3.0; p=0.018). Broader cfDNA NGS panel and CTC-derived ER/HER2 staining results will be presented at the meeting. Conclusions: Among patients receiving A±F, PFS varied based on the PIK3CAmutation status. cfDNA and CTC assessment PreC1 and EOC1 provided complementary information. Further validation in larger datasets is needed. 1. JAMA Oncol 2023;9:815. Clinical trial information: NCT02860000 .

Abstract Ovarian cancer is the most fatal of all female reproductive cancers due to its late detection, aggressive nature, and propensity to develop resistance to current standards of care. Despite initial responses to first line platinum- and paclitaxel-based chemotherapies, &gt;70% of ovarian cancers recur with increasingly resistant disease. Targeted therapies, including PARP-inhibitors (PARPi) and VEGF-inhibitors, are being included in the front-line setting; and while they have improved progression-free survival in a subset of patients, they have thus far failed to improve overall survival in patients without specific genetic signatures. With the broad uptake of PARPi, PARPi-resistance is becoming common. Once resistance develops, nearly all patients die of their disease and there is an obvious need to elucidate the basis for chemotherapy- and PARPi-resistance and devise novel approaches to overcome resistance. Using a small molecule drug screen, we identified lestaurtinib, a known tyrosine kinase inhibitor as a potent inhibitor of treatment naive and highly advanced models of ovarian cancer. Lestaurtinib was originally developed to target FLT3, a protein not expressed in any ovarian cancer models we’ve tested. To identify lestaurtinib targets, we preformed phospho-proteomics in parallel with RNAseq and discovered the JAK/STAT signally pathway as a top hit. We confirmed that essential components of JAK/STAT signaling are constitutively activated in chemotherapy- and PARPi-resistant cells, suggesting that induction of this pathway may drive resistance. Importantly, we found that lestaurtinib-sensitive patient-derived-xenografts (PDXs) exhibited higher expression of multiple JAK/STAT signaling pathway members compared to PDXs that did not respond to lestaurtinib. Genetic knockdown of STAT1 and STAT3 via siRNA, or knockout via CRISPR/Cas9, resulted in significant growth inhibition of sensitive and resistant models, confirming their importance in maintaining cell viability and progression. Combining lestuartinib with standard-of-care cisplatin or olaparib, a PARPi, was shown to be synergistic in multiple models, indicating that pharmacological inhibition of JAK/STAT signaling has the potential to alleviate drug resistance. Finally, we assessed a panel of JAK/STAT inhibitors that interfere at various levels of the pathway and found profound differences in their ability to block STAT-phosphorylation and inhibit cell proliferation. Thus, it is critical to further define the mechanisms by which these drugs function to selectively suppress JAK/STAT signaling in order to elucidate specific nodes of the pathway that drive ovarian cancer progression. Ongoing studies are aimed at identifying the unique roles of STAT1 and STAT3, the specific contributions of their tyrosine and serine phosphorylation, and the relevance of up-/down-stream mediators of the JAK/STAT pathway on ovarian cancer cell growth with the goal of informing future clinical trials. Citation Format: Esther Rodman, Michael Emch, Archit Bajaj, Xiaojia Tang, Xiaonan Hou, Scott Kaufmann, Krishna Kalari, John Weroha, John Hawse. Exploiting JAK/STAT signaling to inhibit highly advanced and resistant forms of ovarian cancer [abstract]. In: Proceedings of the AACR Special Conference on Ovarian Cancer; 2023 Oct 5-7; Boston, Massachusetts. Philadelphia (PA): AACR; Cancer Res 2024;84(5 Suppl_2):Abstract nr B002.

Abstract Background: Dysregulated Notch signalling contributes to breast cancer development and progression, but validated tools to measure the level of Notch signalling in breast cancer subtypes and in response to systemic therapy are largely lacking. A transcriptomic signature of Notch signalling would thus be warranted, and in this report, we have established such a classifier. Methods: To generate the signature, we first identified Notch-regulated genes from six basal-like breast cancer cell lines subjected to elevated and reduced Notch signalling by culturing on immobilized Notch ligand Jagged1 or blockade of Notch by g-secretase inhibitors, respectively. From this cadre of Notch-regulated genes, we developed candidate transcriptomic signatures that were trained on a breast cancer patient dataset (the TCGA-BRCA cohort) and a broader breast cancer cell line cohort and sought to validate in independent data sets. Results: An optimal 20-gene transcriptomic signature was selected. We validated the signature on two independent patient datasets (METABRIC and Oslo2) and it showed an improved coherence score and tumour specificity compared with previously published signatures. Furthermore, the signature score was particularly high for basal-like breast cancer, indicating an enhanced level of Notch signalling in this subtype. The signature score was increased after neoadjuvant treatment in the PROMIX and BEAUTY patient cohorts, and a lower signature score generally correlated with better clinical outcome. Conclusions: The 20-gene transcriptional signature has the potential to better stratify patients and to evaluate the response of future Notch-based therapies for breast cancer. Citation Format: Eike-Benjamin Braune, Felix Geist, Xiaojia Tang, Krishna Kalari, Judy Boughey, Liewei Wang, Roberto A. Leon-Ferre, Antonino B. D’Assoro, James N. Ingle, Matthew P. Goetz, Kang Wang, Theodoros Foukakis, Anita Seshire, Dirk Wienke, Urban Lendahl. Identification of a Notch transcriptomic signature for breast cancer [abstract]. In: Proceedings of the AACR-NCI-EORTC Virtual International Conference on Molecular Targets and Cancer Therapeutics; 2023 Oct 11-15; Boston, MA. Philadelphia (PA): AACR; Mol Cancer Ther 2023;22(12 Suppl):Abstract nr B013.

Abstract Background: Breast cancer is the most frequently diagnosed cancer and the leading cause of cancer-related death among women worldwide. Triple negative breast cancer (TNBC) represents 15-20% of all breast cancers and is an aggressive subtype. Chemotherapy-based treatments remain the standard of care for TNBC. Unfortunately, chemotherapy resistance is common, and for these patients, outcomes are poor and alternative treatment strategies remain an unmet need. circRNAs are a newly identified class of noncoding RNA molecules with covalently closed circular structures. An increasing number of recent studies including ours have indicated that circRNAs play crucial roles in regulating tumor development and chemoresistance. However, the role of circRNAs in the process of chemotherapy resistance and TNBC progression is not clear. Materials and Methods: As a first step towards identifying circRNAs that participate in the development of chemoresistance in TNBC cells, and to determine if targeting such circRNAs is a novel and efficacious therapeutic strategy, doxorubicin-resistant (Doxo-R), paclitaxel-resistant (PTX-R), and double-resistant (DP-R) cell lines were generated from MDA-MB-231. Human circRNA microarrays were utilized to profile the expression of approximately 14,000 known circRNAs in normal breast tissue, matched patient-derived xenografts (PDX) generated prior to and following neoadjuvant chemotherapy (NAC), and TNBC chemosensitive and chemoresistant cell lines. Top hits were validated using RT-PCR. Results: circRNA microarray profiling identified 429 and 310 transcripts differentially expressed in doxorubicin and paclitaxel resistant cells, respectively, compared to parental chemosensitive cell lines (|FC| ≥ 1.5; p value &lt; 0.05). In comparison to pre-NAC derived xenografts, 1,396 circRNAs were dysregulated among post-NAC PDX models. Further, three circRNAs (hsa_circ_001388, hsa_circ_104652, and hsa_circ_061260) were upregulated in Doxo-R, PTX-R, and post-NAC PDX samples compared to their respective controls. Among these three circRNAs, hsa_circ_001388 (also known as circNSD2) was the only transcript also predicted to be translated into a novel and uncharacterized protein given the presence of a high confidence translation initiation site and IRES sequence. Ongoing studies are aimed at determining the role of circNSD2 protein in breast cancer carcinogenesis, progression, and response to standard of care chemotherapeutics and the mechanistic process by which this protein functions. Additionally, the efficacy of specifically targeting this circRNA as a novel therapeutic approach is being explored. Conclusions: Increasing knowledge of the important functions of circRNAs underlying drug resistance will provide new opportunities for developing efficacious therapeutic strategies and prognostic/predictive biomarkers for TNBC. Citation Format: Xiyin Wang, Michael J. Emch, Xiaojia Tang, Jia Yu, Krishna R. Kalari, Liewei Wang, Matthew P. Goetz, John R. Hawse. The role of circular RNAs in triple negative breast cancer and chemotherapy resistance [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2023; Part 1 (Regular and Invited Abstracts); 2023 Apr 14-19; Orlando, FL. Philadelphia (PA): AACR; Cancer Res 2023;83(7_Suppl):Abstract nr 1689.