Abstract Background: HER2-positive breast cancer (HER2+) accounts for approximately 20% of all breast cancers. Major advances in HER2-targeted therapy have improved outcomes; however, substantial room for response improvement remains. Molecular subtyping by BluePrint (BP) stratifies HER2+ disease into HER2 and Luminal types. While BP-HER2 tumors achieve up to 78% pathologic complete response (pCR) with standard therapy, BP-Luminal tumors exhibit persistently low pCR rates (<15%). To address this unmet need, we profiled pretreatment molecular features distinguishing responders from non-responders across all HER2+ tumors and within BP subtypes to identify druggable pathways for rational combination or repurposing strategies. Methods: Baseline microarray profiles from 305 pretreatment HER2+ tumors (87 BP-Luminal, 218 BP-HER2) enrolled across five investigational agents or standard of care in the I-SPY2 trial were analyzed using Gene Set Variation Analysis (GSVA) across 2,265 canonical pathway gene sets. Differential pathway enrichment was assessed using linear models adjusting for treatment arm and false-discovery rate. Two comparisons were performed: (1) pCR vs. no pCR overall and within subtypes, and (2) BP-HER2 vs. BP-Luminal. Shared response- and subtype-specific pathways were grouped by functional similarity and cross-referenced with drug-target databases to identify FDA-approved or investigational compounds. Results: Across HER2+ tumors and within BP-HER2, we identified 42 pathways enriched in non-responders that were also elevated in BP-Luminal relative to BP-HER2, suggesting a luminal-linked resistance program even in non-luminal tumors. These pathways converged into eight metabolic and signaling themes. Growth-factor/RTK bypass (PI3K/AKT) and DNA repair & oxidative stress defense were strongly upregulated in non-responders within BP-HER2, revealing actionable nodes involving PI3K/AKT (alpelisib, capivasertib), IGF1R (linsitinib), MET (crizotinib, capmatinib), and DNA repair (PARP inhibitors). Notably, metabolic rewiring and lipid homeostasis—targetable by vismodegib and sonidegib—were upregulated in non-responders across both BP subtypes, reflecting a shared metabolic vulnerability. Conclusions: Luminal biology-linked transcriptional programs may persist within the HER2+ BP-HER2 subtype and contribute to resistance. BP-HER2 non-responders exhibit coordinated activation of RTK-bypass and DNA repair pathways, exposing therapeutic vulnerabilities targetable by existing agents. Furthermore, targeting lipid metabolic reprogramming alongside anti-HER2 therapy may enhance efficacy and overcome resistance across both BP subtypes. Our future directions include testing these drug combinations in patient-derived HER2+ organoid models. Citation Format: Kingsley V. Chow, Tam Binh Bui, Denise M. Wolf, Annuska Glas, Zheyun Xu, Gillian L. Hirst, I-SPY2 investigators, Amy Clark, Julia Wulfkhule, Angie DeMichele, Emanuel Frank Petricoin, Laura J. Esserman, Jennifer Rosenbluth, Laura van 't Veer, Rosalyn W. Sayaman. Integrative pathway analysis of I-SPY2 HER2+ breast cancers reveals drug-repurposing opportunities [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 2994.

Poor therapeutic response in subsets of breast cancer (BC) patients poses an ongoing challenge. Here, we present a biomarker-guided characterization of 40 patient-derived BC organoids, with the aim of modeling resistant disease with greater fidelity and developing an in vitro system grounded in clinical data for testing alternative treatment strategies. We utilize patient data from the I-SPY2 clinical trial (NCT01042379) to develop predictive models of response to a range of therapies, using only organoid-detectable biomarkers as input, and validate a model predicting response to veliparib-platinum chemotherapy (VP) in triple-negative BC (TNBC) organoids. A drug screen in VP-resistant TNBC organoids reveals combination treatments that overcome resistance to cisplatin, including pro-apoptotic therapies. Another class of hits, HSP90 inhibitors, links organoid drug sensitivity to improved recurrence-free survival in a biomarker-defined patient subset. These findings establish organoid-based functional modeling as a bridge between clinical biomarkers and precision treatment strategies in breast cancer.

610 Background: The adverse event (AE) landscape in oncology is changing due to the introduction of immunotherapy and antibody drug conjugates. These AEs come with both short and long-term symptoms that significantly impact patient quality of life. Monitoring for early onset of symptoms could optimize therapy for a particular patient, maximizing potential efficacy while mitigating toxicity. It is also possible that some toxicities are directly associated with drug sensitivity. We sought to identify symptoms associated with pathologic complete response (pCR) using patient-reported outcomes (PROs) in early-stage high-risk breast cancer patients. Methods: Our study population included 288 stage II/III high-risk breast cancer patients enrolled on the I-SPY2 trial from 2021-2024, who received novel neoadjuvant therapies ± standard paclitaxel. pCR was assigned if tumor was absent in breast and nodes at surgery following neoadjuvant treatment. Patients (n = 288, pCR rate = 29%, 89% administered immunotherapy) were sent electronic PROs. 33 patient-reported AEs were measured using NCI's Patient Reported Outcomes - Common Terminology Criteria for Adverse Events (PRO-CTCAE). Each symptom was evaluated using severity, frequency, and interference on a Likert Scale. Presence of early PRO symptoms (cycles 1-3 of treatment) were binarized (at least one of moderate or greater), and odds ratios were computed with pCR as outcome. To assess whether higher grade AEs were enriched in patients that achieved a pCR, we also performed the Wilcoxan rank sum test using maximum (worst) symptom severity. Results: Of 288 patients included in our analysis (median age = 48 years, range = 20-78), 203 (70.5%) were White, 17 (5.9%) were Asian, 33 (11.5%) were Black or African American, and 35 (12.2%) were Hispanic. PRO analysis revealed that patients that had moderate to severe muscle pain (27% vs 10% OR = 3.15, p < 0.05), joint pain (22% vs 8% OR = 3.23, p < 0.05), headache (27% vs 12.5% OR = 2.59, p < 0.05), or mouth/throat sores (16% vs 5% OR = 3.56, p < 0.05) within weeks 1-3 had higher odds of achieving a pCR. When we looked at maximum severity between weeks 1-3, patients that achieved a PCR had higher grade muscle pain (p = 0.04), heart palpitations (p = 0.035), and significantly lower grade numbness and tingling (p = 0.002). Beyond 6 weeks, associations were weaker or insignificant. Conclusions: Our study utilizes an analysis framework that was able to determine sentinel symptoms such as muscle and joint pain, mouth/throat sores and palpitations as early as weeks 1-3 associated with increased efficacy. This may suggest an early immune reaction in patients that eventually respond favorably to treatment. Our work can help provide earlier proactive monitoring to mitigate toxicities, treatment redirection if needed, and a potential symptom-based early understanding to personalize treatment efficacy. A similar analysis is underway to predict immune related AEs. Clinical trial information: NCT01042379 .

e12755 Background: Pathologic complete response (pCR) is an early endpoint in neoadjuvant breast cancer considered in accelerated approval decisions by the FDA. However trial-level odds ratios of pCR and mature hazard ratios of disease free survival are only weakly associated in meta-analyses, resulting in approvals being delayed until survival evidence becomes available. We used data from the I-SPY 2 trial to develop early endpoints with better trial level correlations to support accelerated approvals. Methods: The I-SPY 2 trial treated 2117 patients with early stage high-risk breast cancer across 26 regimens between 2010 and 2022. When we separate the regimens into groups by subtype, there were 47 total experimental groups (19 HR+HER2-, 19 HR-HER2- and 9 HER2+). Each group is matched to concurrent control group enrolled within 90 days of the enrollment period, forming a experimental/control comparison. Treatment groups with less than 10 patients or insufficient follow-up were excluded, leaving 27 comparisons across 1577 patients. Hazard ratios for event free survival and negative log odds ratios for various early endpoints (e,g, pCR, RCB Class 0 or1) are computed for each experimental/control comparison. Novel early endpoints, including a binary subtype-specific RCB cutoffs, were developed on a training subset, with 5 comparisons held out for validation. The weighted trial-level correlation (R) was calculated overall and in HRHER2 subgroups. Subtype specific RCB cutoffs were chosen to optimize the overall and within subgroup weighted correlation, while maintaining C indices > .70 for individual survival association. More sophisticated models with RCB index as a continuous variable that incorporated baseline T stage, nodal status and response predictive subtypes were also explored. Results: Overall weighted correlations (R) for PCR, RCB 0/1 and Subtype Specific RCB cutoffs are shown in Table 1. Optimal cutoffs of RCB of 1.1 for HER2+, 1.7 for HR-HER2-, and 2.6 for HR+HER2- were identified giving a correlation of .70 overall in the training set and .54 in training and validation combined. The C index (C) was greater than 0.8 for all endpoints indicating good individual association. Preliminary results using more sophisticated models will be presented. Conclusions: Subtype-specific RCB cutoffs show a modest but clear improvement as an early endpoint over RCB-0/1 and pCR while maintaining a strong patient-level association with survival. External validation efforts are underway. Integrative approaches using continuous RCB improve trial level correlation and will be presented. Trial-level weighted correlation and individual C indices across all 3 subtypes and within each HRHER2 subtype separately. Measure OverallR,C HER2+R,C HR+HER2-R,C HR-HER2-R,C PCR .23, .84 .31, .80 .13, .84 .18, .87 RCB 0-1 .31, .83 .37, .80 .38, .85 .16, .85 Subtype Specific Cutpoint .54, .80 .37, .77 .40, .75 .54, .87 R = correlation; C = C index.

Background:Circulating tumor DNA (ctDNA) is an emerging biomarker of treatment response and recurrence risk, while residual cancer burden (RCB) after neoadjuvant treatment (NAT) is a well-established risk factor for distant recurrence. Here, we examined the association between high ctDNA concentration at diagnosis and risk of distant recurrence after neoadjuvant treatment (NAT), in the context of RCB. Methods:The study included 712 patients with high-risk breast cancer in the neoadjuvant I-SPY2 trial. Tumor- informed ctDNA test results at diagnosis were used to stratify patients into ctDNA-negative and ctDNA-positive groups. For this analysis, the ctDNA-positive group was divided into tertiles (low, intermediate, high) based on ctDNA concentration reported as mean tumor molecules per mL [MTM/mL] of plasma. Correlations between MTM/mL at diagnosis and ctDNA dynamics during NAT, residual cancer burden (RCB), and distant recurrence-free survival (DRFS) were examined across all subtypes. Results:In all subtypes, high ctDNA concentration at diagnosis was associated with worse DRFS, whereas low ctDNA concentration or ctDNA-negative status was associated with improved DRFS, even with high tumor burden after NAT (RCB-II/RCB-III). We also found that patients with high ctDNA concentration, regardless of subtype, were less likely to experience early ctDNA clearance; however, those who did had a significantly higher likelihood of achieving a favorable response (RCB-0/RCB-I) than those with late or no ctDNA clearance. Furthermore, across all subtypes, patients with early ctDNA clearance, including those with substantial residual cancer (RCB-II/RCB-III) after NAT, had improved DRFS, irrespective of the ctDNA concentration at diagnosis. Conclusions:Across all subtypes, pathologic response and ctDNA clearance reduce the risk of distant recurrence associated with high ctDNA concentration at diagnosis. ctDNA concentration at diagnosis and ctDNA clearance dynamics during NAT may facilitate the prediction of treatment response and further stratify the risk of metastatic recurrence in non-responders. Trial Registration: NCT01042379.

Dynamic biomarkers of therapy response are critical for precision oncology but often rely on serial tissue biopsies, which are invasive and not always feasible. In contrast, peripheral blood offers a minimally invasive, dynamic window into the evolving systemic immune landscape. Leveraging this, we performed RNA sequencing on 546 peripheral blood samples from 160 patients with high-risk stage II/III human epidermal growth factor receptor 2 (HER2)-negative breast cancer treated with either chemotherapy alone or in combination with immunotherapy (chemoimmunotherapy). Our analysis uncovered immune correlates of tumor subtype and treatment response. For example, samples from patients with triple-negative breast cancer exhibited elevated T cell receptor clonality and robust immune activation profiles. Among patients receiving chemoimmunotherapy, early responders demonstrated high baseline T cell receptor diversity, followed by rapid clonal expansion and activation of T cells after just one treatment cycle. We developed a multiparametric peripheral immune biomarker that integrated baseline and early on-treatment features to predict response to pembrolizumab, which was successfully validated in an independent cohort of 59 patients with breast cancer treated with neoadjuvant dostarlimab. These findings reveal the potential of blood-based immune monitoring to predict immunotherapy benefit, offering an accessible tool for tailoring treatment strategies in breast cancer.

618 Background: In neoadjuvant cancer trials, early endpoints that predict treatment effect on survival identify promising agents early and support accelerated regulatory approval. However, binary endpoints like pathologic complete response inadequately characterize the full distribution of residual disease in breast cancer, while promising continuous biomarkers like MRI derived functional tumor volume (FTV) are associated with survival outcomes but not widely collected. Using a Bayesian hierarchical model, long-term treatment effects on distant recurrence free survival (DRFS) can be predicted from continuous MRI-derived functional tumor volume (FTV) in the I-SPY 2 platform trial. Methods: I-SPY 2 treated 2117 patients from 2010-2022 (12 weeks of paclitaxel ± experimental agent followed by 4 cycles of doxorubicin + cyclophosphamide), and 1,859 underwent dynamic contrast-enhanced MRI at baseline and after 12 weeks of neoadjuvant therapy. MRI-derived functional tumor volume provided volumetric quantification of dynamic tissue enhancement. ΔFTV was defined as the ratio of 12-week to baseline FTV. A Bayesian joint hierarchical model (brms) fit treatment effects on ΔFTV and DRFS for each treatment regimen by HR/HER2 subtype, controlling for clinical nodal status, clinical T stage, grade, and calendar year. Arms with < 8 subjects are excluded. Performance was assessed using cross-validation, predicting DRFS treatment effects in one held out fold at a time from the learned ΔFTV-DRFS association in the remaining data, then comparing predicted to actual DRFS treatment effect. Sensitivity analyses on priors will be presented. Results: Across 1753 patients and 45 treatment–subtype combinations, the estimated treatment effects on ΔFTV and DRFS were highly correlated (posterior correlation -0.91; 95% CrI -1.00 to -0.23). Predicted DRFS treatment effect from ΔFTV demonstrated strong concordance with actual DRFS treatment effects (Pearson r = 0.94 in TNBC; 0.97 HER2+; 0.80 HR+HER2-). The top 5 treatment-subtype regimens ranked by predicted and actual DRFS were identical. 20 regimens predicted to have > 70% probability of DRFS benefit over subtype specific controls showed DRFS improvement, yielding 100% specificity and 69% sensitivity at this decision threshold. Conclusions: We demonstrate internally validated prediction of neoadjuvant treatment effect on DRFS from MRI-derived change in functional tumor volume in the I-SPY 2 trial of high-risk early breast cancer. This suggests continuous imaging measures capture a range of response to therapy while Bayesian approaches can be effective for predicting treatment effects. This encourages collecting MRI biomarkers in trials to facilitate validation as early endpoints supporting decisions in screening platform trials as well as regulatory accelerated approval.

PURPOSE:Neoadjuvant immunotherapy (IO) has become the standard of care for early-stage triple-negative breast cancer (TNBC), but not yet for other subtypes. We previously developed a clinical-grade mRNA-based immune classifier (ImPrint) predicting response to IO that is now being used in I-SPY2.2 as part of the response predictive subtypes. We report the performance of ImPrint in hormone receptor-positive and human epidermal growth factor receptor 2-negative (HR+HER2-) patients from five IO arms. METHODS:A total of 204 HR+HER2- (MammaPrint high-risk) patients from five IO arms (anti-PD-1, anti-PD-L1/poly [ADP-ribose] polymerase inhibitor combination, anti-PD-1/toll-like receptor 9 dual-IO combination, and anti-PD-1 ± lymphocyte activation gene 3 dual-IO combination) and 191 patients from the chemotherapy-only control arm were included in this analysis. Patients were classified as ImPrint+ (likely sensitive) versus ImPrint- (likely resistant), using pretreatment mRNA. Performance of ImPrint for predicting pathologic complete response (pCR) to IO-containing arms was characterized and compared with tumor grade (III), MammaPrint (ultra) High2 risk (MP2), and estrogen receptor (ER)-low (ER ≤ 10%). RESULTS:Overall, the pCR rate across the five IO arms was 33%. 26% of HR+HER2- patients were ImPrint+, and pCR rates with IO were 75% in ImPrint+ versus 17% in ImPrint-, with the highest pCR rate >90% in a dual-IO arm. In the control arm, pCR rates were 33% in ImPrint+ and 8% in ImPrint-. Tumor grade (III), MP2, and ER-low showed pCR rates in IO of 45%, 56%, and 63%, respectively, with lower pCR odds ratios (OR < 7.5) compared with ImPrint (OR = 14.5). CONCLUSION:Using an accurate selection strategy, HR+HER2- patients could achieve pCR rates similar to what is seen with best neoadjuvant therapies in TNBC and HER2+ (ie, pCR rate >65%-70%). ImPrint, an Food and Drug Administration IDE-enabled assay, may represent a way to identify HR+HER2- patients for IO that best balances likely benefit versus risk of serious immune-related adverse events.

Introduction: The tumor immune microenvironment (TME) is a strong prognostic factor in breast cancer with higher proportions of tumor infiltrating lymphocytes associated with longer survival. Features of the breast TME vary by race/ethnicity and are linked to disparities in outcomes. In 2021, we performed the first comprehensive investigation of the effect of germline variants on immune traits associated with survival. Leveraging pan-cancer data from The Cancer Genome Atlas (TCGA), we identified 33 heritable immune traits and >1500 SNPs associated with the TME. We hypothesize that racial/ethnic disparities in cancer outcomes may be partially driven by germline SNPs that differentially mediate the TME. Here, we assess the differential expression of heritable immune traits and prevalence of immune-associated germline SNPs across genetic ancestry groups in TCGA breast cancer cohort. Lastly, we assess whether these heritable TME features alone are robust predictors of genetic ancestry. Methods: Utilizing expression-based signatures of immune traits and germline SNP data in TCGA breast cancers (n=986), we investigated the distribution of heritable immune traits and immune-associated germline variants across genetic ancestry groups (European EUR=762, African AFR=163, Asian ASN=45, Amerindian AMR n=16). Differential expression of heritable traits across genetic ancestries were assessed using Kruskal-Wallis test. Proportions of genotypes across genetic ancestries were compared using multiple pair-wise Fisher’s exact tests. P-values were adjusted for multiple testing, and significance defined at adj. p<0.05. Multi-class prediction of EUR, AFR and ASN genetic ancestry was performed using random forest machine learning models. Cross-validation was performed in 75% of samples and 25% used for independent validation. Immune traits and/or germline SNPs were used as features and model performance was evaluated using mean balanced accuracy (Acc). Results: Expression of 33% of heritable immune traits varied by genetic ancestry in breast cancer (adj. p<0.05). Enrichment of eosinophils, natural killer, Th17 and T central memory cells were lower while expression of Interferon, PD-1, and antigen-presentation signatures were higher in AFR vs. EUR. Enrichment of Th2 cells was higher in ASN vs. EUR, while cytotoxic cells was higher in AMR vs. EUR. Moreover, 31% of immune-associated SNPs had significant differences in genotype proportions between at least two ancestry groups. Prevalence of 454 SNPs varied between AFR and EUR and included SNPs associated with CD8 T cells and Interferon signaling. Immune-associated germline SNPs (Acc=0.96) and immune traits+germline SNPs (Acc=0.99) were robust predictors of EUR, AFR and ASN ancestry. Conclusion: Prevalence of heritable components of the TME varied across genetic ancestries. By identifying which features of the TME that drive treatment responses are mediated by heritable germline factors, we can further elucidate mechanisms that underlie disparities in treatment outcomes. Further validation of these findings is underway. Citation Format: Rosalyn W Sayaman, Denise M Wolf, Christina Yau, Vesteinn Thorsson, Mohamad Saad, Donglei Hu, Scott Huntsman, Davide Bedognetti, Michael J Campbell, Jennifer Rosenbluth, Elad Ziv. Differential prevalence of immune-associated germline variants across genetic ancestry groups may partially underlie racial and ethnic disparities in breast cancer outcomes [abstract]. In: Proceedings of the AACR Special Conference in Cancer Research: Functional and Genomic Precision Medicine in Cancer: Different Perspectives, Common Goals; 2025 Mar 11-13; Boston, MA. Philadelphia (PA): AACR; Cancer Res 2025;85(5 Suppl):Abstract nr B042.

Early-stage breast cancers resistant to neoadjuvant therapy (NAT), characterized by high residual cancer burden (RCB) after treatment, have an increased risk of metastatic recurrence. Here, we show that circulating tumor DNA (ctDNA) detected using a tumor-informed test (1) can improve risk stratification of patients with NAT-resistant tumors (RCB-II/RCB-III) and (2) predict response to NAT. Stratification using ctDNA status at pretreatment or post-NAT and ctDNA dynamics identified NAT-resistant tumors with a significantly decreased risk of metastatic recurrence. ctDNA clearance as early as week 3 across receptor subtypes predicted favorable responses to NAT, including immunotherapies. Interestingly, less than a fifth of patients with NAT-resistant tumors were ctDNA-positive post-NAT. Serial mutation profiling of NAT-resistant tumors revealed that patient-specific ctDNA assay variants remained detectable over time, including in tumors of patients ctDNA-negative post-NAT. Refining risk stratification for NAT-resistant tumors using ctDNA and understanding ctDNA shedding in these tumors could guide treatment decisions to prevent or delay metastatic recurrence.

PURPOSE:Pretreatment specimens from patients treated on the I-SPY2 neoadjuvant breast cancer trial were studied to identify prespecified biomarkers associated with response to the regimen of paclitaxel, the anti-type I insulin-like growth factor receptor (IGF-1R) antibody ganitumab, and metformin (PGM) followed by doxorubicin and cyclophosphamide (AC) compared with control therapy (paclitaxel followed by AC). The primary endpoint of this trial is pathologic complete response (pCR). EXPERIMENTAL DESIGN:One hundred six patients treated with PGM and 119 contemporary controls were evaluated using laser capture microdissection and reverse-phase protein array to evaluate 32 prespecified potential predictive biomarkers in the IGF-1R pathway and 109 additional exploratory endpoints. RESULTS:Total levels of IGF-1R were poorly correlated with phosphorylated IGF-1R/insulin receptor (IR). Higher levels of phosphorylated IGF-1R/IR were associated with an increased likelihood of obtaining pCR, especially in the hormone receptor (HR)-positive subgroup. Markers of immune response also showed an association with pCR but differed between HR+ and HR- subgroups. In HR- tumors, phospho-STAT1 Y701 and low levels of phospho-p27 associated with pCR. These relationships were not observed in patients treated with control chemotherapy. CONCLUSIONS:Activation status of IGF-1R/IR associated with increased pCR to PGM in HR+ breast cancers. Immune activation markers were also associated with response in HR+ and HR- subgroups. Thus, IGF-1R may directly regulate tumor biology and associate with immune response to therapy.

587 Background: GES predictive of response to therapy across multiple breast cancer subtypes are commercially available or in development. Deep learning models can predict GES from digital histology, and may serve as a lower-cost alternative immediately available at the time of biopsy. Methods: Transformer-based models trained to predict 38 distinct breast cancer signatures from pathology (all with Pearson correlation > 0.5 versus true GES) were previously developed using cases from The Cancer Genome Atlas. These models were applied to digital H&E from pre-treatment biopsies from HER2- cases treated with CT or CT + immunotherapy (IO) from the ISPY2 trial. The histology-derived GES most predictive of pCR in ISPY2 (as per area under the ROC curve [AUROC]) was tested in two external neoadjuvant cohorts - a subset of a trial from Yale of durvalumab + CT (NCT02489448) with TIL annotations, and patients receiving standard of care CT at University of Chicago. AUROC significance was assessed with 1000x bootstrapping, with Benjamini Hochberg correction applied in ISPY2 to account for testing multiple GES models. Tertiles of predicted expression calculated in ISPY2 defined groups with low, medium, and high likelihood of pCR; these cutoffs were tested in the external cohorts. Results: Accuracy for pCR prediction was tested in 578 patients from seven arms of ISPY2 – with breakdown by treatment and hormone receptor (HR) status shown in Table. A histology model for a GES defined by estrogen regulated genes (Oh et al, JCO 2006) – including proliferation, apoptosis, and interferon-response genes – predicted pCR with the highest AUROC (0.794) in ISPY2, and outperformed a logistic regression fit on grade, HR status, and tumor / nodal stage (AUROC 0.705, p for comparison 0.0001). Tertiles of predicted expression for this GES (computed in ISPY2) identified groups with low / high pCR rates which were robust to treatment, HR status, and consistent in validation cohorts (Table). This digital signature (AUROC 0.737) compared favorably to pathologist TIL annotation (AUROC 0.664) from the external Yale cohort. An explainability tool demonstrated that patterns of lymphocytic infiltrate and poor differentiation contributed to high signature predictions from histology. Conclusions: A digital histology-derived GES consistently identifies patients at low / high likelihood of pCR with neoadjuvant CT or CT + IO, and may improve treatment personalization. Subgroup n AUROC p % pCR (low expression) % pCR (mid expression) % pCR (high expression) ISPY2 579 0.794 2 x 10 -28 7.6 26.7 58.6 ISPY2, CT + IO 459 0.810 3 x 10 -26 8.0 28.2 64.1 ISPY2, CT only 120 0.726 0.001 6.4 20.0 36.8 ISPY2, HR- 239 0.704 4 x 10 -7 14.8 29.2 58.4 ISPY2, HR+ 340 0.817 1 x 10 -15 6.5 24.8 59.0 UChicago HR- 151 0.746 3 x 10 -7 11.1 27.3 55.1 UChicago HR+ 63 0.847 1 x 10 -5 5.5 12.0 70.0 Yale HR- 41 0.737 0.005 0.0 50.0 61.5

Motivation: Previously, we developed MRI-based models for the prediction of pathologic complete response (pCR) using an initial cohort of 990 patients enrolled in I-SPY. The purpose of this study is to validate the performance of the MRI model using an independent patient cohort from I-SPY 2. Goal(s): The goal is to test the robustness of the MRI-based models. Approach: Area under the receiver operating characteristic curve (AUC), PPV, and sensitivity for pCR prediction was used to evaluate performance in the subsequent cohort. Results: Overall, the sensitivity and PPV were 72% and 57%, slightly higher than the values evaluated in the initial 990 cohort. Impact: This is the first study of MRI-based predictive models that were developed and validated using two separate large cohorts from a multicenter neoadjuvant chemotherapy clinical trial.

PURPOSE:The MammaPrint (MP) prognostic assay categorizes breast cancers into high- and low-risk subgroups, and the high-risk group can be further subdivided into high-1 (MP-H1), and very high-risk high-2 (MP-H2). The aim of this analysis was to assess clinical and molecular differences between the hormone receptor-positive (HR+)/HER2-negative MP-H1, -H2, and triple-negative (TN) MP-H1 and -H2 cancers. EXPERIMENTAL DESIGN:Pretreatment gene expression data from 742 HER2-negative breast cancers enrolled in the I-SPY2 neoadjuvant trial were used. Prognostic risk categories were assigned using the MP assay. Transcriptional similarities across the four receptor and prognostic groups were assessed using principal component analyses and by identifying differentially expressed genes. We also examined pathologic complete response rates and event-free survivals by risk group. RESULTS:Principal component analysis showed that HR+/MP-H2 tumors clustered with TN/MP-H2 cancers. Only 125 genes showed differential expression between the HR+/MP-H2 and TN/MP-H2 cancers, whereas 1,465 genes were differentially expressed between HR+/MP-H2 and -H1. Gene set analysis revealed similarly high expression of cell cycle, DNA repair, and immune infiltration-related pathways in HR+/MP-H2 and TN/MP-H2 cancers. HR+/MP-H2 cancers also showed low estrogen receptor-related gene expression. Pathologic complete response rates were similarly high in TN/MP-H2 and HR+/MP-H2 cancers (42% vs. 30.5%; P = 0.11), and MP-H2 cancers with residual cancer had similarly poor event-free survival regardless of estrogen receptor status. CONCLUSIONS:In conclusion, HR+/MP-H2 cancers closely resemble TN breast cancers in transcriptional and clinical features and benefit from similar treatment strategies.

Background: Functional tumor volume (FTV), measured from dynamic contrast-enhanced MRI, is an imaging biomarker that can predict treatment response in breast cancer patients undergoing neoadjuvant chemotherapy (NAC). The FTV-based predictive model, combined with core biopsy, informed treatment decisions of recommending patients with excellent responses to proceed to surgery early in a large NAC clinical trial. Methods: In this retrospective study, we constructed models using FTV measurements. We analyzed performance tradeoffs when a probability threshold was used to identify excellent responders through the prediction of pathology complete response (pCR). Individual models were developed within cohorts defined by the hormone receptor and human epidermal growth factor receptor 2 (HR/HER2) subtype. Results: A total of 814 patients enrolled in the I-SPY 2 trial between 2010 and 2016 were included with a mean age of 49 years (range: 24 to 77). Among these patients, 289 (36%) achieved pCR. The area under the ROC curve (AUC) ranged from 0.68 to 0.74 for individual HR/HER2 subtypes. When probability thresholds were chosen based on minimum positive predictive value (PPV) levels of 50%, 70%, and 90%, the PPV-sensitivity tradeoff varied among subtypes. The highest sensitivities (100%, 87%, 45%) were found in the HR−/HER2+ sub-cohort for probability thresholds of 0, 0.62, and 0.72; followed by the triple-negative sub-cohort (98%, 52%, 4%) at thresholds of 0.13, 0.58, and 0.67; and HR+/HER2+ (78%, 16%, 8%) at thresholds of 0.34, 0.57, and 0.60. The lowest sensitivities (20%, 0%, 0%) occurred in the HR+/HER2− sub-cohort. Conclusions: Predictive models developed using imaging biomarkers, alongside clinically validated probability thresholds, can be incorporated into decision-making for precision oncology.

Poor therapeutic response in subsets of breast cancer (BC) patients poses an ongoing challenge. Here, we present a biomarker-guided characterization of 44 patient-derived BC organoids, with the aim of modeling resistant disease with greater fidelity and developing an in-vitro system grounded in clinical data for testing alternative treatment strategies. We utilized patient transcriptomic and outcome data from the I-SPY2 clinical trial to develop predictive models of response to a range of therapies, using only organoid-detectable biomarkers as input. A model predicting response to veliparib-platinum chemotherapy (VP) in triple-negative BC (TNBC) was validated in organoids, showing that in vitro drug responses matched predictions from the patient data-derived model. A drug screen in VP-resistant TNBC organoids identified combination treatments that overcame resistance to cisplatin, including pro-apoptotic therapies. This demonstrates that gene expression-based resistance models derived from patient data can be successfully modeled in organoids that can then be used for therapeutic evaluation.