Abstract Background: The PREDIX HER2 trial compared standard neoadjuvant therapy with 6 cycles of docetaxel, trastuzumab, and pertuzumab (DTP), versus 6 cycles of trastuzumab emtansine (T-DM1) in 197 patients with HER2-positive breast cancer. There was no difference in pathologic complete response (pCR) rate and event-free survival (EFS) between the two treatments (Hatschek, JAMA Oncology 2021). Here we systematically evaluate the prognostic and predictive molecular biomarkers during neoadjuvant HER2-targeting therapy. Methods: Longitudinal fresh-frozen tissue biopsies (pretreatment (n=194) and after 2 cycles (n=167)) and surgical specimens (n=126) were collected and sequenced by RNA-sequencing (RNA-seq). Differential gene expression (DGE) analyses were conducted using zero-inflated negative binomial mixed model, and P-values were adjusted by the Benjamini-Hochberg method. Potential prognostic and predictive markers including PAM50 intrinsic subtype, cancer hallmark signature (n=50) score, tumor infiltrating lymphocyte fraction (n=9), and immune/stromal score were calculated based on normalized count data. The correlations between above biomarkers and pCR and EFS were analyzed using multivariate logistic and Cox regressions, respectively. We integrated the least absolute shrinkage and selection operator (LASSO) regression and bootstrapping algorithm (iteration=10,000, nfold=5) to choose best predictive features. Results: Downregulation of proliferation-related and extracellular matrix pathways (DGEs with false discovery rate (FDR) < 0.05, |log fold change|>1) was observed throughout treatment in both arms. DTP resulted in early (on-treatment vs baseline) inflammatory (IL-17, TGF-β, TNF signaling pathways) and immune (B cell, NK cell and cytokine signaling) responses, whereas late responses occurred in the T-DM1 arm (post-treatment vs on-treatment). Interestingly, a rebound of HER2 and PI3K-AKT signaling was observed within residual disease after T-DM1. Immune and stromal scores showed similar kinetics between the two treatment arms, with increase after first two cycles and later decrease to baseline levels. PAM50 intrinsic subtype at baseline was independently associated with pCR and EFS after adjusting for treatment arm, tumor size, lymph node status, hormone receptor status, and Ki-67. Luminal A (pCR rate, 11.1%, odds ratio (OR), 0.64, 95% confidence interval (CI): 0.49 to 0.84, P=0.001), Luminal B (pCR rate, 20.0%, OR, 0.63, 95% CI: 0.53 to 0.76, P< 0.001) and basal-like (pCR rate, 33.3%, OR, 0.70, 95% CI: 0.53 to 0.93, P=0.02) subtypes had lower pCR rates compared to HER2-enriched subtype (, 68.9%). Patients with basal-like tumors at baseline had worse EFS than those with HER2-enriched tumors (hazard ratio (HR), 4.66, 95% CI, 1.28 to 16.90, P= 0.02). Moreover, pair-wise analyses revealed that patients with HER2-enriched tumors at baseline switching to other PAM50 subtypes after two cycles, had improved pCR (OR, 1.54, 95% CI: 1.31 to 1.80, P< 0.001) and EFS (HR), 0.26, 95% CI, 0.08 to 0.83, P= 0.02) than the remaining patients. Machine learning based analyses identified best biomarkers for pCR (hallmark estrogen response early, hallmark androgen response, hallmark PI3K AKT mTOR signaling) and EFS (NK and Treg cells fraction, hallmark apical surface). Patients with higher hallmark apical surface score could benefit from T-DM1 (HRT-DM1 vs. DTP = 0.13, 95% CI, 0.02 to 0.79, P=0.03) and vice versa (HRT-DM1 vs. DTP = 5.02, 95% CI, 1.06 to 23.76, P=0.04). Conclusion: This study sheds light on how the tumor transcriptome evolves under anti-HER2 therapy and potentially provides prognostic and predictive biomarkers for standard chemotherapy with dual HER2 blockade versus monotherapy with an antibody-drug conjugate. Further investigations evaluating the spatial and single-cell heterogeneity of HER2-positive BC are ongoing. Citation Format: Kang Wang, Yajing Zhu, Ioannis Zerdes, Alexios Matikas, Emmanouil Sifakis, Dimitrios Salgkamis, Judith Bjöhle, Ellinor Elinder, Sara Margolin, Ana Bosch Campos, Gyula Pekar, Henrik Lindman, Aglaia Schiza, Zakaria Einbeigi, Jamila Adra, Anne Andersson, Lena Carlsson, Ann Charlotte Dreifaldt, Erika Isaksson-Friman, Susanne Agartz, Hemming Johansson, Mats Hellström, Edward Azavedo, Johan Hartman, Jonas Bergh, Thomas Hatschek, Theodoros Foukakis. Correlative and longitudinal transcriptomic profiling predicts patient outcomes and the efficacy of neoadjuvant HER2-targeted treatments in the randomized PREDIX HER2 trial [abstract]. In: Proceedings of the 2022 San Antonio Breast Cancer Symposium; 2022 Dec 6-10; San Antonio, TX. Philadelphia (PA): AACR; Cancer Res 2023;83(5 Suppl):Abstract nr P5-02-20.