4D Path's AI Platform Predicts Chemotherapy Response in Triple-Negative Breast Cancer Using Standard Biopsy Images
核心洞察
4D Path (搜索)'s QPOR™ platform successfully predicted treatment response in early-stage triple-negative breast cancer (搜索) patients using only digitized baseline biopsy images, representing the first physics-based computational biomarker to predict outcomes from pre-treatment tissue alone.
The platform demonstrated 9-fold increased response odds in paclitaxel-treated patients (OR 9.17, p=0.009; AUC 0.74) by analyzing routine H&E slides without requiring clinical or molecular inputs.
Published in the Journal of the National Cancer (搜索) Institute, the study establishes a new framework for baseline digital image-based prediction of chemotherapy response and provides a scalable, cost-effective alternative to conventional AI models.
4D Path (搜索) announced that its Q-Plasia OncoReader (QPOR™) platform has successfully predicted treatment response in patients with early-stage triple-negative breast cancer (搜索) (TNBC (搜索)) using digitized baseline biopsy images. The analysis, conducted within the Translational Breast Cancer Research Consortium (搜索) (TBCRC) 030 clinical trial, represents the first demonstration of a statistical physics and tumor biology-based computational biomarker predicting post-treatment outcomes from pre-treatment tissue alone, without any clinical or molecular inputs.
Results published in the Journal of the National Cancer (搜索) Institute establish a new framework for baseline digital image-based prediction of chemotherapy response and lay the groundwork for future chemo-immunotherapy applications for early-stage TNBC (搜索). The multi-institutional collaboration was led by investigators from Dana-Farber Cancer Institute and included researchers from Harvard Medical School, Brigham and Women's Hospital, 4D Path (搜索) Inc., and other leading academic centers.
Breakthrough Predictive Performance
The TBCRC 030 trial previously evaluated Myriad Genetics (搜索)' HRD biomarker as a predictor of response to either cisplatin or paclitaxel in TNBC (搜索). In this follow-up analysis, investigators applied 4D Path (搜索)'s QPOR™ algorithm to pre-treatment core biopsies to evaluate tumor-infiltrating lymphocytes (TILs) and computational immune-cell-cycle biomarkers (CmbI, also known as the complex immune response index CIRi), in addition to visual examination by a pathologist. All assessments were blinded and independently performed on digitized hematoxylin and eosin (H&E) slides.
The platform demonstrated remarkable predictive accuracy, with CmbI predicting RCB 0/1 in paclitaxel-treated patients with an odds ratio of 9.17 (p=0.009; AUC 0.74) using H&E slides only. Notably, CA-TILs and IHI did not achieve similar predictive performance. Both visual and computational assessments correlated with response, underscoring the prognostic and predictive role of immune infiltration in TNBC (搜索).
In the BRCA1 (搜索)/2-proficient cohort, visually assessed TILs and CmbI associated with response overall, with the signal localizing specifically to paclitaxel treatment rather than cisplatin.
Physics-Based Computational Approach
QPOR's CmbI is designed to capture the baseline complex immune response as a predictor of post-operative residual cancer (搜索) burden (RCB) after neoadjuvant chemotherapy. Computed directly from H&E slides, it integrates the tumor's proliferative state and the spatial heterogeneity of immune infiltrates in conjunction with a cell cycle G1/S (搜索) deregulation signature.
"These findings show the complex immune response—a key ingredient of tumor-response biology—can be quantitatively captured directly from routine biopsy slides using first-principles computation as a predictive biomarker of neoadjuvant chemotherapy response, and that G1/S (搜索) cell-cycle deregulation signatures can be leveraged for a comprehensive quantification of the complex immune response," said Dr. Satabhisa Mukhopadhyay, Co-founder and Chief Scientific Officer of 4D Path (搜索).
Clinical Validation and Generalizability
The predictive capacity observed with paclitaxel in TBCRC 030 was also demonstrated in TBCRC 031, involving germline BRCA-mutated, HER2 (搜索)-negative breast cancer (搜索), where CmbI predicted response to neoadjuvant doxorubicin-cyclophosphamide (AC) and cisplatin monotherapy. These observations across different patient populations support the potential generalizability of the assay.
"This is a prospective evaluation of a computational biomarker derived from standard histology that, much like expert visual assessment of TILs, predicts post-treatment response," said Dr. Erica Mayer, principal investigator of TBCRC 030 and director of breast cancer (搜索) clinical research at Dana-Farber Cancer Institute. "Developing reproducible predictive signatures from clinical trial pre-treatment biopsies could ultimately improve selection of therapies to optimize patient care."
Technological Advantages
The computational analysis produced a quantitative, fully deterministic, and reproducible biomarker predictive of treatment response—by design, intrinsically free of observer variability and machine-learning bias. By relying solely on H&E images, the QPOR™ platform offers a scalable, cost-effective, and interpretable alternative to conventional AI-driven models that require large, annotated datasets.
The study underscores the potential of computational pathology and oncology to complement traditional visual methods, expand biomarker access globally, and facilitate standardized clinical trial workflows. This approach provides a foundation for treatment predictive modeling in chemotherapy, potentially applicable to chemoimmunotherapy, and supports the use of physics-based algorithms as complementary tools for patient selection, treatment de-escalation, or early therapy switching decisions.