A gradient tree boosting and network propagation derived pan-cancer survival network of the tumor microenvironment

Abstract

Predicting cancer survival from molecular data is an important aspect of biomedical research because it allows quantifying patient risks and thus individualizing therapy. We introduce XGBoost tree ensemble learning to predict survival from transcriptome data of 8,024 patients from 25 different cancer types and show highly competitive performance with state-of-the-art methods. To further improve plausibility of the machine learning approach we conducted two additional steps. In the first step, we applied pan-cancer training and showed that it substantially improves prognosis compared with cancer subtype-specific training. In the second step, we applied network propagation and inferred a pan-cancer survival network consisting of 103 genes. This network highlights cross-cohort features and is predictive for the tumor microenvironment and immune status of the patients. Our work demonstrates that pan-cancer learning combined with network propagation generalizes over multiple cancer types and identifies biologically plausible features that can serve as biomarkers for monitoring cancer survival.

Gradient tree boosting and network propagation for the identification of pan-cancer survival networks

“…Download and documentation is available from https://git-scm.com/ . CRITICAL: The indicated software and package versions were used in ( Thedinga and Herwig, 2022 ). Other versions than the ones indicated here were not tested.…”

“…For these patients, only the sample with the lexicographically highest TCGA barcode is kept. Note: For the analyses in ( Thedinga and Herwig, 2022 ) and shown in the Figures of this protocol, TCGA data was downloaded using manifest files and the GDC Data Transfer Tool. This can lead to a different ordering of samples as compared to downloading the data with TCGAbiolinks, which results in different training and test data splits when training the XGBoost survival prediction model.…”

“…In case of multiple samples corresponding to the same case ID, the first sample was kept. In case you want to reproduce the exact results from ( Thedinga and Herwig, 2022 ) and shown in the Figures of this protocol, use the GDC Data Transfer Tool to download the TCGA data based on the manifest files provided in the “data” directory at https://github.molgen.mpg.de/thedinga/xgb_survival_network . Keep only samples with sample type annotation “Primary Tumor” and “Primary Blood Derived Cancer – Peripheral Blood” and for multiple samples corresponding to the same case ID, keep the first sample.…”

“…This protocol is based on TCGA transcriptome data of 8,024 patients from 25 cancer types but can easily be adapted to cancer patient data from other sources. For complete details on the use and execution of this protocol, please refer to Thedinga and Herwig (2022) .…”

“…For complete details on the use and execution of this protocol, please refer to Thedinga and Herwig (2022) .…”