Contrastive self-supervised clustering of scRNA-seq data

Abstract

Background Single-cell RNA sequencing (scRNA-seq) has emerged has a main strategy to study transcriptional activity at the cellular level. Clustering analysis is routinely performed on scRNA-seq data to explore, recognize or discover underlying cell identities. The high dimensionality of scRNA-seq data and its significant sparsity accentuated by frequent dropout events, introducing false zero count observations, make the clustering analysis computationally challenging. Even though multiple scRNA-seq clustering techniques have been proposed, there is no consensus on the best performing approach. On a parallel research track, self-supervised contrastive learning recently achieved state-of-the-art results on images clustering and, subsequently, image classification. Results We propose contrastive-sc, a new unsupervised learning method for scRNA-seq data that perform cell clustering. The method consists of two consecutive phases: first, an artificial neural network learns an embedding for each cell through a representation training phase. The embedding is then clustered in the second phase with a general clustering algorithm (i.e. KMeans or Leiden community detection). The proposed representation training phase is a new adaptation of the self-supervised contrastive learning framework, initially proposed for image processing, to scRNA-seq data. contrastive-sc has been compared with ten state-of-the-art techniques. A broad experimental study has been conducted on both simulated and real-world datasets, assessing multiple external and internal clustering performance metrics (i.e. ARI, NMI, Silhouette, Calinski scores). Our experimental analysis shows that constastive-sc compares favorably with state-of-the-art methods on both simulated and real-world datasets. Conclusion On average, our method identifies well-defined clusters in close agreement with ground truth annotations. Our method is computationally efficient, being fast to train and having a limited memory footprint. contrastive-sc maintains good performance when only a fraction of input cells is provided and is robust to changes in hyperparameters or network architecture. The decoupling between the creation of the embedding and the clustering phase allows the flexibility to choose a suitable clustering algorithm (i.e. KMeans when the number of expected clusters is known, Leiden otherwise) or to integrate the embedding with other existing techniques.

Miscell: An efficient self-supervised learning approach for dissecting single-cell transcriptome

“…Momentum contrastive self-supervised learning achieved comparable performance in visual representation learning of images as compared with supervised representation learning ( Chen et al., 2020b ; He et al., 2019 ). As compared with a similar method proposed by Ciotran and colleagues ( Ciortan and Defrance, 2021 ) used a network of 3 linear layers as feature encoder, we observed that Miscell achieved better performance ( Figure S6 ). This is probably because of better representation capacity of the feature encoder used by Miscell.…”

Comparing representations of biological data learned with different AI paradigms, augmenting and cropping strategies

“…Most metrics, however, require the ground truth labelling, which were not available in this study. Besides, the clustering itself can be approached in many different ways, using the classical or the newly developed deep-learning based algorithms (Ciortan and Defrance, 2021). In this study, we only intended to fairly compare clustering results, obtained under identical conditions (same algorithm, grid search parameters, evaluation metrics, etc.)…”

“…Finally, self-supervision has been successfully applied to cell segmentation, annotation and clustering (Lu et al, 2019;Santos-Pata et al, 2021). Most recently, a self-supervised contrastive learning framework has been proposed by Ciortan and Defrance (2021) to learn representations of scRNA-seq data. The authors follow the idea of SimCLR (Chen et al, 2020) and show state-of-the-art (SOTA) performance on clustering task.…”

Self-supervised clustering of mass spectrometry imaging data using contrastive learning

“…Following its success in computer vision, this strategy has been adopted in several applications in other research fields including classification of electrocardiograms 31 and clustering of scRNA-seq data. 32 It has been demonstrated that the development of modality-specific data augmentation is critical to the performance of models trained using contrastive learning.…”