CONE: COntext-specific Network Embedding via Contextualized Graph Attention

Renming Liu, Hao Yuan, Kayla Johnson, Arjun Krishnan
Proceedings of the 19th Machine Learning in Computational Biology meeting, PMLR 261:53-71, 2024.

Abstract

Human gene interaction networks, commonly known as interactomes, encode genes’ functional relationships, which are invaluable knowledge for translational medical research and the mechanistic understanding of complex human diseases. Advanced network embedding techniques have inspired recent efforts to identify novel human disease-associated genes using canonical interactome embeddings. However, a pivotal challenge persists as many complex diseases manifest in specific biological contexts, such as tissues or cell types, while many existing interactomes do not encapsulate such information. Here, we propose CONE (\url{https://github.com/krishnanlab/cone}), a versatile approach to generate context-specific embeddings from any context-free interactomes. The core component of CONE consists of a graph attention network with contextual conditioning, which is trained in a noise-contrastive fashion using contextualized interactome random walks localized around contextual genes. We demonstrate the strong performance of CONE embeddings in identifying disease-associated genes when using known associated biological contexts to the diseases. Furthermore, our approach offers new insights into the biological contexts associated with human diseases.

Cite this Paper

BibTeX
@InProceedings{pmlr-v261-liu24a, title = {CONE: COntext-specific Network Embedding via Contextualized Graph Attention}, author = {Liu, Renming and Yuan, Hao and Johnson, Kayla and Krishnan, Arjun}, booktitle = {Proceedings of the 19th Machine Learning in Computational Biology meeting}, pages = {53--71}, year = {2024}, editor = {Knowles, David A and Mostafavi, Sara}, volume = {261}, series = {Proceedings of Machine Learning Research}, month = {05--06 Sep}, publisher = {PMLR}, pdf = {https://raw.githubusercontent.com/mlresearch/v261/main/assets/liu24a/liu24a.pdf}, url = {https://proceedings.mlr.press/v261/liu24a.html}, abstract = {Human gene interaction networks, commonly known as interactomes, encode genes’ functional relationships, which are invaluable knowledge for translational medical research and the mechanistic understanding of complex human diseases. Advanced network embedding techniques have inspired recent efforts to identify novel human disease-associated genes using canonical interactome embeddings. However, a pivotal challenge persists as many complex diseases manifest in specific biological contexts, such as tissues or cell types, while many existing interactomes do not encapsulate such information. Here, we propose CONE (\url{https://github.com/krishnanlab/cone}), a versatile approach to generate context-specific embeddings from any context-free interactomes. The core component of CONE consists of a graph attention network with contextual conditioning, which is trained in a noise-contrastive fashion using contextualized interactome random walks localized around contextual genes. We demonstrate the strong performance of CONE embeddings in identifying disease-associated genes when using known associated biological contexts to the diseases. Furthermore, our approach offers new insights into the biological contexts associated with human diseases.} }
Endnote
%0 Conference Paper %T CONE: COntext-specific Network Embedding via Contextualized Graph Attention %A Renming Liu %A Hao Yuan %A Kayla Johnson %A Arjun Krishnan %B Proceedings of the 19th Machine Learning in Computational Biology meeting %C Proceedings of Machine Learning Research %D 2024 %E David A Knowles %E Sara Mostafavi %F pmlr-v261-liu24a %I PMLR %P 53--71 %U https://proceedings.mlr.press/v261/liu24a.html %V 261 %X Human gene interaction networks, commonly known as interactomes, encode genes’ functional relationships, which are invaluable knowledge for translational medical research and the mechanistic understanding of complex human diseases. Advanced network embedding techniques have inspired recent efforts to identify novel human disease-associated genes using canonical interactome embeddings. However, a pivotal challenge persists as many complex diseases manifest in specific biological contexts, such as tissues or cell types, while many existing interactomes do not encapsulate such information. Here, we propose CONE (\url{https://github.com/krishnanlab/cone}), a versatile approach to generate context-specific embeddings from any context-free interactomes. The core component of CONE consists of a graph attention network with contextual conditioning, which is trained in a noise-contrastive fashion using contextualized interactome random walks localized around contextual genes. We demonstrate the strong performance of CONE embeddings in identifying disease-associated genes when using known associated biological contexts to the diseases. Furthermore, our approach offers new insights into the biological contexts associated with human diseases.
APA
Liu, R., Yuan, H., Johnson, K. & Krishnan, A.. (2024). CONE: COntext-specific Network Embedding via Contextualized Graph Attention. Proceedings of the 19th Machine Learning in Computational Biology meeting, in Proceedings of Machine Learning Research 261:53-71 Available from https://proceedings.mlr.press/v261/liu24a.html.

Related Material