iMAP: integration of multiple single-cell datasets by adversarial paired transfer networks

Wang, Dongfang; Hou, Siyu; Zhang, Lei; Wang, Xiliang; Liu, Baolin; Zhang, Zemin

doi:10.1186/s13059-021-02280-8

Cited by 43 publications

(72 citation statements)

References 44 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Our GAN-GMHI framework consists of three stages, constructing a dataset containing phenotype and batch information for all samples, and then GAN guiding the batch effect correction of raw data, the corrected datasets are output as the training data set for GMHI prediction ( Supplementary Figure 1 ). The batch effect removal method of iMAP (Wang, et al, 2021), a GAN method previously applied on single-cell RNA-Seq data, was adapted for batch effect removal in this study. It is worth noting that the datasets to be batch-corrected by GAN must be classified based on the phenotype first, and the sub-data sets of each phenotype are regrouped according to the batch.…”

Section: Methodsmentioning

confidence: 99%

GAN-GMHI: Generative Adversarial Network for high discrimination power in microbiome-based disease prediction

Xie

Zha

et al. 2021

Preprint

View full text Add to dashboard Cite

Summary: Gut microbiome-based health index (GMHI) has been applied with success, while the discrimination powers of GMHI varied for different diseases, limiting its utility on a broad spectrum of diseases. In this work, a Generative Adversarial Network (GAN) model is proposed to improve the discrimination power of GMHI. Built based on the batch corrected data through GAN (https://github.com/HUST-NingKang-Lab/GAN-GMHI), GAN-GMHI has largely reduced the batch effects, and profoundly improved the performance for distinguishing healthy individuals and different diseases. GAN-GMHI has provided results to support the strong association of gut microbiome and diseases, and indicated a more accurate venue towards microbiome-based disease monitoring. Availability and implementation: GAN-GMHI is publicly available on GitHub: https://github.com/HUST-NingKang-Lab/GAN-GMHI. Contact: ningkang@hust.edu.cn

show abstract

Section: Methodsmentioning

confidence: 99%

GAN-GMHI: Generative Adversarial Network for high discrimination power in microbiome-based disease prediction

Xie

Zha

et al. 2021

Preprint

View full text Add to dashboard Cite

show abstract

“…In single-cell genomics, a number of approaches for removing inter-experimental variability from data have been developed [17][18][19][20][21][22][23][24]. Two such methods are Harmony [25] and scGen [26].…”

Section: Related Workmentioning

confidence: 99%

Removing Inter-Experimental Variability from Functional Data in Systems Neuroscience

Gonschorek

Höfling

Szatko

et al. 2021

Preprint

View full text Add to dashboard Cite

Integrating data from multiple experiments is common practice in systems neuroscience but it requires inter-experimental variability to be negligible compared to the biological signal of interest. This requirement is rarely fulfilled; systematic changes between experiments can drastically affect the outcome of complex analysis pipelines. Modern machine learning approaches designed to adapt models across multiple data domains offer flexible ways of removing inter-experimental variability where classical statistical methods often fail. While applications of these methods have been mostly limited to single-cell genomics, in this work, we develop a theoretical framework for domain adaptation in systems neuroscience. We implement this in an adversarial optimization scheme that removes inter-experimental variability while preserving the biological signal. We compare our method to previous approaches on a large-scale dataset of two-photon imaging recordings of retinal bipolar cell responses to visual stimuli. This dataset provides a unique benchmark as it contains biological signal from well-defined cell types that is obscured by large inter-experimental variability. In a supervised setting, we compare the generalization performance of cell type classifiers across experiments, which we validate with anatomical cell type distributions from electron microscopy data. In an unsupervised setting, we remove inter-experimental variability from data which can then be fed into arbitrary downstream analyses. In both settings, we find that our method achieves the best trade-off between removing inter-experimental variability and preserving biological signal. Thus, we offer a flexible approach to remove inter-experimental variability and integrate datasets across experiments in systems neuroscience. Code available at https://github.com/eulerlab/rave.

show abstract

“…The two batches of the CRC dataset were sequenced based on different methods, and the two methods had their own characteristics. The 10X data (10x original data) are sparser but have higher throughput in terms of cell numbers, while the Smart-seq2 data measure more accurate gene expression levels [13,37,38].…”

Section: Biological Characteristics Discovery Based On Awganmentioning

confidence: 99%

“…The strategy we choose to improve the performance of our model is based on both the visualization result and quantitative assessments (especially ASW rate and LISI rate). We also refer from the stage2 of iMAP's structure for network design [13].…”

Section: Model Adjustment and Evaluationmentioning

confidence: 99%

“…Even with the many methods available, the task of batch correction still presents many difficulties. The most crucial point is to find a way to perfectly mix cells from the same cell type across different batches and to be able to discern batch-specific cell types [13]. The existing methods have different focuses on the above two points, but the different focuses of some methods can lead to poor results sometimes after batch effect removal (see our discussion part).…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

ResPAN: a powerful batch correction model for scRNA-seq data through residual adversarial networks

Liu

Wang

Zhao

2021

Preprint

View full text Add to dashboard Cite

With the advancement of technology, we can generate and access large-scale, high dimensional and diverse genomics data, especially through single-cell RNA sequencing (scRNA-seq). However, integrative downstream analysis from multiple scRNA-seq datasets remains challenging due to batch effects. In this paper, we focus on scRNA-seq data integration and propose a new deep learning framework based on Wasserstein Generative Adversarial Network (WGAN) combined with an attention mechanism to reduce the differences among batches. We also discuss the limitations of the existing methods and demonstrate the advantages of our new model from both theoretical and practical aspects, advocating the use of deep learning in genomics research.

show abstract

iMAP: integration of multiple single-cell datasets by adversarial paired transfer networks

Cited by 43 publications

References 44 publications

GAN-GMHI: Generative Adversarial Network for high discrimination power in microbiome-based disease prediction

GAN-GMHI: Generative Adversarial Network for high discrimination power in microbiome-based disease prediction

Removing Inter-Experimental Variability from Functional Data in Systems Neuroscience

ResPAN: a powerful batch correction model for scRNA-seq data through residual adversarial networks

Contact Info

Product

Resources

About