journal contribution
posted on 2024-05-02, 08:05 authored by Katie E. Blise, Shamilene Sivagnanam, Courtney B. Betts, Konjit Betre, Nell Kirchberger, Benjamin J. Tate, Emma E. Furth, Andressa Dias Costa, Jonathan A. Nowak, Brian M. Wolpin, Robert H. Vonderheide, Jeremy Goecks, Lisa M. Coussens, Katelyn T. Byrne Supplementary Figure S3. A. SHAP plot showing the top 30 features driving the IA model. Features are ordered
on the y-axis such that those with a larger impact on the model’s predictions appear at the top of the SHAP plot.
SHAP values are shown on the x-axis, with a value of zero (center) indicating no impact on the model, and
negative or positive SHAP values predicting long DFS or short DFS, respectively. Red or blue dots indicate
presence or absence, respectively, of the corresponding feature in tissues. B. Box plot showing feature values for
each of the top 15 features for the model derived from IA regions of the αCD40 cohort split by DFS group (n = 30
regions from short DFS patients per feature; n = 13 regions from long DFS patients per feature). Each dot
represents the log10+1 normalized feature value for one tissue region, which was inputted into the classifier model. Boxes = Q1 to Q3; whiskers = smallest and largest
datapoints within 1.5*IQR +/- Q3/Q1; solid line = median. Mann–Whitney U-test used to determine
statistical significance. P-values corrected using the Benjamini–Hochberg procedure. *, P ≤ 0.05; **, P ≤ 0.01;
***, P ≤ 0.001.
Funding
National Cancer Institute (NCI)
United States Department of Health and Human Services
Find out more...Dana-Farber Cancer Institute Hale Family Center for Pancreatic Cancer Research
Lustgarten Foundation Dedicated Laboratory Program
Parker Institute for Cancer Immunotherapy (PICI)
Brenden-Colson Center for Pancreatic Care
Robert L. Fine Cancer Research Foundation
Prospect Creek Foundation
Knight Cancer Institute, Oregon Health and Science University (KCI)
History
ARTICLE ABSTRACT
Tumor molecular data sets are becoming increasingly complex, making it nearly impossible for humans alone to effectively analyze them. Here, we demonstrate the power of using machine learning (ML) to analyze a single-cell, spatial, and highly multiplexed proteomic data set from human pancreatic cancer and reveal underlying biological mechanisms that may contribute to clinical outcomes. We designed a multiplex immunohistochemistry antibody panel to compare T-cell functionality and spatial localization in resected tumors from treatment-naïve patients with localized pancreatic ductal adenocarcinoma (PDAC) with resected tumors from a second cohort of patients treated with neoadjuvant agonistic CD40 (anti-CD40) monoclonal antibody therapy. In total, nearly 2.5 million cells from 306 tissue regions collected from 29 patients across both cohorts were assayed, and over 1,000 tumor microenvironment (TME) features were quantified. We then trained ML models to accurately predict anti-CD40 treatment status and disease-free survival (DFS) following anti-CD40 therapy based on TME features. Through downstream interpretation of the ML models’ predictions, we found anti-CD40 therapy reduced canonical aspects of T-cell exhaustion within the TME, as compared with treatment-naïve TMEs. Using automated clustering approaches, we found improved DFS following anti-CD40 therapy correlated with an increased presence of CD44+CD4+ Th1 cells located specifically within cellular neighborhoods characterized by increased T-cell proliferation, antigen experience, and cytotoxicity in immune aggregates. Overall, our results demonstrate the utility of ML in molecular cancer immunology applications, highlight the impact of anti-CD40 therapy on T cells within the TME, and identify potential candidate biomarkers of DFS for anti-CD40–treated patients with PDAC.
