Analysis of computational tumor-infiltrating lymphocytes in breast cancer from the results of the TIGER challenge

M. Rijthoven, W. Aswolinskiy, L. Tessier, R. Salgado, J. van der Laak, F. Ciompi, T. consortium, M. van Rijthoven, M. Balkenhol, J. Bogaerts, D. Drubay, L. Blesa, D. Peeters, E. Stovgaard, A. L\aenkholm, H. Haynes, L. Craciun, D. Larsimont, M. Amgad, L. Cooper, C. de Kock, V. Dechering, J. Lotz, N. Weiss, M. van Bockstal, C. Galant, E. Lips, H. Horlings, J. Wesseling, L. Mulder, S. van den Belt, K. Weber, P. Jank, C. Denkert, E. Munari, G. Bogina, C. Russ, A. Lemm, S. Loi, J. Dixon-Douglas, S. Michiels, R. Donders, S. Maurits, M. Groeneveld, A. Mickan, J. Meakin, B. van Ginneken, H. Joensuu, M. Fan, D. Lee, J. Ye, K. Byun, J. Kim, S. Xu, Z. Ji, F. Xie, J. Kuang, X. Chen, L. Chen, A. Arab, W. Chen, V. Garcia, N. Petrick, B. Gallas, A. Tsakiroglou, R. Byers, M. Fergie, V. Ramanathan, A. Martel, A. Shephard, S. Ahmed Raza, M. Jahanifar, N. Rajpoot, S. Cho, D. Kim, H. Jang, C. Park and K. Kim

Nature Communications 2026.

DOI

Abstract

Tumor-infiltrating lymphocytes (TILs) is a recognized prognostic biomarker in breast cancer. However, poor interobserver agreement and limited reproducibility highlight the need for computational approaches. Despite advances, adoption of computational models has been hindered by lack of standardized methods and robust benchmarks. To address this, we launched TIGER, an international competition to build open-source computational TILs (cTILs) models. Here, we present a multi-centric analysis of cTILs methods on resections and biopsies from 3,708 human epidermal growth factor receptor 2-positive (HER2+) or triple-negative breast cancers (TNBC) from clinical practice and phase 3 trials. We report benchmarks on image analysis performance, show strong agreement of cTILs with pathologists, and demonstrate positive association of cTILs with neoadjuvant therapy response in HER2+, superior to visually scored TILs. We also show that cTILs add independent prognostic information to clinical variables in TNBC resections. Data, methods and benchmarks are publicly available:

https://tiger.grand-challenge.org/

.