Explainable Deep Neural Network for Lung Squamous Cell Carcinoma Survival Analysis by Integrating Genomic and Clinical Data

Xudan Zhou, Qinglin Yang, Yuxin Zhang, Yanyan Hou, Changlong Chen, Guohui Ma, Jin Luo, Wei Shu
Proceedings of 2025 2nd International Conference on Machine Learning and Intelligent Computing, PMLR 278:734-746, 2025.

Abstract

we utilized explainable deep learning methodologies to elucidate critical genes and prospective biomarkers correlated with the prognosis of Lung Squamous Cell Carcinoma (LUSC). Transcriptomic data were systematically acquired from the TCGA repository and underwent comprehensive differential expression profiling to identify candidate genes warranting in-depth exploration. We developed Cox-PASNet, a pathway-aware deep learning model designed to predict survival outcomes in lung squamous cell carcinoma (LUSC) by integrating multi-modal data, including clinical variables, transcriptomic profiles, and curated biological pathways. The model demonstrated robust performance, achieving an AUC of 0.73 in stratifying patients into long- and short-term survival groups. Beyond predictive accuracy, Cox-PASNet offers interpretable insights into key molecular pathways, facilitating the discovery of novel prognostic biomarkers (CCDC181, B2M, BTD, C1orf112, ANAPC7) and their related biological pathways (regulation of cell cycle, DNA repair, cytoskeletal dynamics, tumor microenvironment, and metastasis) associated with LUSC survival. The significance of these genes was validated using external datasets and clinical indicators. Notably, members of the CCDC family were particularly important, with many found to enhance tumor cell proliferation. Elevated expression levels of CCDC proteins demonstrated a significant correlation with adverse clinical outcomes, including diminished overall survival rates and unfavorable prognosis. In summary, through interpretable deep learning and bioinformatics approaches, we identified several relevant genes, with CCDC genes being closely linked to LUSC survival.

Cite this Paper

BibTeX
@InProceedings{pmlr-v278-zhou25a, title = {Explainable Deep Neural Network for Lung Squamous Cell Carcinoma Survival Analysis by Integrating Genomic and Clinical Data}, author = {Zhou, Xudan and Yang, Qinglin and Zhang, Yuxin and Hou, Yanyan and Chen, Changlong and Ma, Guohui and Luo, Jin and Shu, Wei}, booktitle = {Proceedings of 2025 2nd International Conference on Machine Learning and Intelligent Computing}, pages = {734--746}, year = {2025}, editor = {Zeng, Nianyin and Pachori, Ram Bilas and Wang, Dongshu}, volume = {278}, series = {Proceedings of Machine Learning Research}, month = {25--27 Apr}, publisher = {PMLR}, pdf = {https://raw.githubusercontent.com/mlresearch/v278/main/assets/zhou25a/zhou25a.pdf}, url = {https://proceedings.mlr.press/v278/zhou25a.html}, abstract = {we utilized explainable deep learning methodologies to elucidate critical genes and prospective biomarkers correlated with the prognosis of Lung Squamous Cell Carcinoma (LUSC). Transcriptomic data were systematically acquired from the TCGA repository and underwent comprehensive differential expression profiling to identify candidate genes warranting in-depth exploration. We developed Cox-PASNet, a pathway-aware deep learning model designed to predict survival outcomes in lung squamous cell carcinoma (LUSC) by integrating multi-modal data, including clinical variables, transcriptomic profiles, and curated biological pathways. The model demonstrated robust performance, achieving an AUC of 0.73 in stratifying patients into long- and short-term survival groups. Beyond predictive accuracy, Cox-PASNet offers interpretable insights into key molecular pathways, facilitating the discovery of novel prognostic biomarkers (CCDC181, B2M, BTD, C1orf112, ANAPC7) and their related biological pathways (regulation of cell cycle, DNA repair, cytoskeletal dynamics, tumor microenvironment, and metastasis) associated with LUSC survival. The significance of these genes was validated using external datasets and clinical indicators. Notably, members of the CCDC family were particularly important, with many found to enhance tumor cell proliferation. Elevated expression levels of CCDC proteins demonstrated a significant correlation with adverse clinical outcomes, including diminished overall survival rates and unfavorable prognosis. In summary, through interpretable deep learning and bioinformatics approaches, we identified several relevant genes, with CCDC genes being closely linked to LUSC survival.} }
Endnote
%0 Conference Paper %T Explainable Deep Neural Network for Lung Squamous Cell Carcinoma Survival Analysis by Integrating Genomic and Clinical Data %A Xudan Zhou %A Qinglin Yang %A Yuxin Zhang %A Yanyan Hou %A Changlong Chen %A Guohui Ma %A Jin Luo %A Wei Shu %B Proceedings of 2025 2nd International Conference on Machine Learning and Intelligent Computing %C Proceedings of Machine Learning Research %D 2025 %E Nianyin Zeng %E Ram Bilas Pachori %E Dongshu Wang %F pmlr-v278-zhou25a %I PMLR %P 734--746 %U https://proceedings.mlr.press/v278/zhou25a.html %V 278 %X we utilized explainable deep learning methodologies to elucidate critical genes and prospective biomarkers correlated with the prognosis of Lung Squamous Cell Carcinoma (LUSC). Transcriptomic data were systematically acquired from the TCGA repository and underwent comprehensive differential expression profiling to identify candidate genes warranting in-depth exploration. We developed Cox-PASNet, a pathway-aware deep learning model designed to predict survival outcomes in lung squamous cell carcinoma (LUSC) by integrating multi-modal data, including clinical variables, transcriptomic profiles, and curated biological pathways. The model demonstrated robust performance, achieving an AUC of 0.73 in stratifying patients into long- and short-term survival groups. Beyond predictive accuracy, Cox-PASNet offers interpretable insights into key molecular pathways, facilitating the discovery of novel prognostic biomarkers (CCDC181, B2M, BTD, C1orf112, ANAPC7) and their related biological pathways (regulation of cell cycle, DNA repair, cytoskeletal dynamics, tumor microenvironment, and metastasis) associated with LUSC survival. The significance of these genes was validated using external datasets and clinical indicators. Notably, members of the CCDC family were particularly important, with many found to enhance tumor cell proliferation. Elevated expression levels of CCDC proteins demonstrated a significant correlation with adverse clinical outcomes, including diminished overall survival rates and unfavorable prognosis. In summary, through interpretable deep learning and bioinformatics approaches, we identified several relevant genes, with CCDC genes being closely linked to LUSC survival.
APA
Zhou, X., Yang, Q., Zhang, Y., Hou, Y., Chen, C., Ma, G., Luo, J. & Shu, W.. (2025). Explainable Deep Neural Network for Lung Squamous Cell Carcinoma Survival Analysis by Integrating Genomic and Clinical Data. Proceedings of 2025 2nd International Conference on Machine Learning and Intelligent Computing, in Proceedings of Machine Learning Research 278:734-746 Available from https://proceedings.mlr.press/v278/zhou25a.html.

Related Material