ACE: Explaining cluster from an adversarial perspective

Yang Young Lu, Timothy C Yu, Giancarlo Bonora, William Stafford Noble
Proceedings of the 38th International Conference on Machine Learning, PMLR 139:7156-7167, 2021.

Abstract

A common workflow in single-cell RNA-seq analysis is to project the data to a latent space, cluster the cells in that space, and identify sets of marker genes that explain the differences among the discovered clusters. A primary drawback to this three-step procedure is that each step is carried out independently, thereby neglecting the effects of the nonlinear embedding and inter-gene dependencies on the selection of marker genes. Here we propose an integrated deep learning framework, Adversarial Clustering Explanation (ACE), that bundles all three steps into a single workflow. The method thus moves away from the notion of "marker genes" to instead identify a panel of explanatory genes. This panel may include genes that are not only enriched but also depleted relative to other cell types, as well as genes that exhibit differences between closely related cell types. Empirically, we demonstrate that ACE is able to identify gene panels that are both highly discriminative and nonredundant, and we demonstrate the applicability of ACE to an image recognition task.

Cite this Paper

BibTeX
@InProceedings{pmlr-v139-lu21e, title = {ACE: Explaining cluster from an adversarial perspective}, author = {Lu, Yang Young and Yu, Timothy C and Bonora, Giancarlo and Noble, William Stafford}, booktitle = {Proceedings of the 38th International Conference on Machine Learning}, pages = {7156--7167}, year = {2021}, editor = {Meila, Marina and Zhang, Tong}, volume = {139}, series = {Proceedings of Machine Learning Research}, month = {18--24 Jul}, publisher = {PMLR}, pdf = {http://proceedings.mlr.press/v139/lu21e/lu21e.pdf}, url = {https://proceedings.mlr.press/v139/lu21e.html}, abstract = {A common workflow in single-cell RNA-seq analysis is to project the data to a latent space, cluster the cells in that space, and identify sets of marker genes that explain the differences among the discovered clusters. A primary drawback to this three-step procedure is that each step is carried out independently, thereby neglecting the effects of the nonlinear embedding and inter-gene dependencies on the selection of marker genes. Here we propose an integrated deep learning framework, Adversarial Clustering Explanation (ACE), that bundles all three steps into a single workflow. The method thus moves away from the notion of "marker genes" to instead identify a panel of explanatory genes. This panel may include genes that are not only enriched but also depleted relative to other cell types, as well as genes that exhibit differences between closely related cell types. Empirically, we demonstrate that ACE is able to identify gene panels that are both highly discriminative and nonredundant, and we demonstrate the applicability of ACE to an image recognition task.} }
Endnote
%0 Conference Paper %T ACE: Explaining cluster from an adversarial perspective %A Yang Young Lu %A Timothy C Yu %A Giancarlo Bonora %A William Stafford Noble %B Proceedings of the 38th International Conference on Machine Learning %C Proceedings of Machine Learning Research %D 2021 %E Marina Meila %E Tong Zhang %F pmlr-v139-lu21e %I PMLR %P 7156--7167 %U https://proceedings.mlr.press/v139/lu21e.html %V 139 %X A common workflow in single-cell RNA-seq analysis is to project the data to a latent space, cluster the cells in that space, and identify sets of marker genes that explain the differences among the discovered clusters. A primary drawback to this three-step procedure is that each step is carried out independently, thereby neglecting the effects of the nonlinear embedding and inter-gene dependencies on the selection of marker genes. Here we propose an integrated deep learning framework, Adversarial Clustering Explanation (ACE), that bundles all three steps into a single workflow. The method thus moves away from the notion of "marker genes" to instead identify a panel of explanatory genes. This panel may include genes that are not only enriched but also depleted relative to other cell types, as well as genes that exhibit differences between closely related cell types. Empirically, we demonstrate that ACE is able to identify gene panels that are both highly discriminative and nonredundant, and we demonstrate the applicability of ACE to an image recognition task.
APA
Lu, Y.Y., Yu, T.C., Bonora, G. & Noble, W.S.. (2021). ACE: Explaining cluster from an adversarial perspective. Proceedings of the 38th International Conference on Machine Learning, in Proceedings of Machine Learning Research 139:7156-7167 Available from https://proceedings.mlr.press/v139/lu21e.html.

Related Material