Speaker: Arianna Rigamonti
June 23rd, 2026 | 2:15 pm
DEIB, PT1 Meeting Room (Bld. 20A)
For further information please contact: Silvia Cascianelli | silvia.cascianelli@polimi.it
Abstract
Tuesday, June 23rd, 2026 at 2:15 pm a new appointment of Data Science Seminars: Bioinformatics focus will take place in DEIB PT1 Meeting Room (Building 20A) organized by the Data Science for Bioinformatics group.The seminar will be held by Arianna Rigamonti, PHD Student in Information Technology, on the following subject: "Longitudinal Neutrophil Subpopulation Dynamics as Biomarkers of Immunotherapy Response in Non-Small Cell Lung Cancer: A Machine Learning Survival Analysis".
Non-small cell lung cancer (NSCLC) accounts for the majority of lung cancer cases worldwide and remains a leading cause of cancer-related mortality. Over the past decade, immune checkpoint inhibitors (ICIs) have transformed the treatment landscape, producing durable responses in a subset of patients. Yet a major unsolved challenge is predicting, at the individual level, who will benefit from immunotherapy and who will not. Current biomarkers, such as PD-L1 expression and Tumor Mutational Burden (TMB), offer only partial discrimination, motivating the search for new predictors of response. Among immune cells that actively shape the tumor microenvironment, low-density neutrophils (LDNs) have recently emerged as key modulators of immunotherapy response. LDNs can be found in functionally distinct subsets that either promote or restrain anti-tumor immunity. Flow cytometry allows quantification of these subpopulations at high resolution, but their longitudinal behavior over the course of treatment has been poorly studied. Within the APOLLO11 trial, a longitudinal study conducted at the Istituto Nazionale dei Tumori of Milan, LDN subpopulation dynamics in NSCLC patients were profiled at three timepoints: treatment start (T0), early follow-up (T1, within the first 50 days), and late follow-up (T2, between 50 and 150 days). The dataset comprises three cohorts: 117 patients receiving immunotherapy alone (IO), 179 receiving immunotherapy combined with chemotherapy (IOct), and 67 receiving targeted therapy (TT). To model survival outcomes (overall survival and progression-free survival), we developed a two-track machine learning pipeline. The first track applies regularized Cox regression, gradient-boosted machines, and random survival forests. The second track uses a recurrent neural network (RNN) to explicitly model the trajectory of immune markers over time. Model performance was evaluated using the C-index within a repeated nested cross-validation framework. SHAP-based feature attribution was then used to identify the most predictive LDN subpopulations and their temporal dynamics. The results of this study demonstrate that longitudinal LDN profiles, captured through routine blood flow cytometry, can serve as dynamic biomarkers of immunotherapy response, enabling early patient stratification and therapeutic adjustments for predicted non-responders.