Conditional out-of-distribution generation for unpaired data using transfer VAE
Open accessThe DOI matches its Crossref record; 11 of the 14 references it cites were found in a registry or an index.
Abstract
Motivation: While generative models have shown great success in sampling high-dimensional samples conditional on low-dimensional descriptors (stroke thickness in MNIST, hair color in CelebA, speaker identity in WaveNet), their generation out-of-distribution poses fundamental problems due to the difficulty of learning compact joint distribution across conditions. The canonical example of the conditional variational autoencoder (CVAE), for instance, does not explicitly relate conditions during training and, hence, has no explicit incentive of learning such a compact representation. Results: We overcome the limitation of the CVAE by matching distributions across conditions using maximum mean discrepancy in the decoder layer that follows the bottleneck. This introduces a strong regularization both for reconstructing samples within the same condition and for transforming samples across conditions, resulting in much improved generalization. As this amount to solving a style-transfer problem, we refer to the model as transfer VAE (trVAE). Benchmarking trVAE on high-dimensional image and single-cell RNA-seq, we demonstrate higher robustness and higher accuracy than existing approaches. We also show qualitatively improved predictions by tackling previously problematic minority classes and multiple conditions in the context of cellular perturbation response to treatment and disease based on high-dimensional single-cell gene expression data. For generic tasks, we improve Pearson correlations of high-dimensional estimated means and variances with their ground truths from 0.89 to 0.97 and 0.75 to 0.87, respectively. We further demonstrate that trVAE learns cell-type-specific responses after perturbation and improves the prediction of most cell-type-specific genes by 65%. Availability and implementation: The trVAE implementation is available via github.com/theislab/trvae. The results of this article can be reproduced via github.com/theislab/trvae_reproducibility.
Overview
Integrity signals
Registry
Cites
Statistics
Verification
Claims the paper makes
Claims other papers make about this one
Each is checked against this paper’s own text: does it say what the citing paper says it does?
In the citing paper
“Multiple distance measures for scRNA-seq have been explored by the single-cell community in recent years, including Wasserstein distance computed in an optimal transport framework22,23, maximum mean discrepancy24, neighborhood-based measures17,18, and energy distance (E-distance)20.”
From scPerturb: harmonized single-cell perturbation data (2024) · cites this paper as [24]
In the citing paper
“Contrary to scGEN, trVAE can address the issue of performing OOD predictions for multiple perturbations within a given dataset.”
From A mini-review on perturbation modelling across single-cell omic modalities (2024) · cites this paper as [53]
In the citing paper
“Aside from CPA, existing methods (Lotfollahi et al, 2020; Russkikh et al, 2020) such as scGen (Lotfollahi et al, 2019) have also been shown capable of predicting single‐cell perturbation responses when the dataset contains no combinatorial treatment or dose‐dependent perturbations.”
From Predicting cellular responses to complex perturbations in high‐throughput screens (2023) · cites this paper as Lotfollahi et al, 2020
Citations
Cites
Advances in neural information processing systems 7
(1996) · Computers & Mathematics with Applications · doi:10.1016/s0898-1221(96)90463-0 · Cited by 14,281
Advances in Neural Information Processing Systems 29
Teymur, O., Zygalakis, K., & Calderhead, B. (2016) · Cited by 13,887
(2012) · MPG.PuRe (Max Planck Society) · doi:10.5555/2188385.2188410 · Cited by 3,603
Deep generative modeling for single-cell transcriptomics
Lopez, R., Regier, J., Cole, M. B., Jordan, M. I., & Yosef, N. (2018) · Nature Methods · doi:10.1038/s41592-018-0229-2 · Cited by 2,530
A single-cell survey of the small intestinal epithelium
Haber, A. L., Biton, M., Rogel, N., Herbst, R. H., Shekhar, K., Smillie, C., Burgin, G., Delorey, T. M., Howitt, M. R., Katz, Y., et al. (2017) · Nature · doi:10.1038/nature24489 · Cited by 1,750
Multiplexed droplet single-cell RNA-sequencing using natural genetic variation
Kang, H. M., Subramaniam, M., Targ, S., Nguyen, M., Maliskova, L., McCarthy, E., Wan, E., Wong, S., Byrnes, L., Lanata, C. M., et al. (2018) · Nature Biotechnology · doi:10.1038/nbt.4042 · Cited by 1,101
Single-cell RNA-seq denoising using a deep count autoencoder
Eraslan, G., Simon, L. M., Mircea, M., Mueller, N. S., & Theis, F. J. (2019) · Nature Communications · doi:10.1038/s41467-018-07931-2 · Cited by 1,084
Self-citationscGen predicts single-cell perturbation responses
Lotfollahi, M., Wolf, F. A., & Theis, F. J. (2019) · Nature Methods · doi:10.1038/s41592-019-0494-8 · Cited by 661
Self-citationMassively multiplex chemical transcriptomics at single-cell resolution
Srivatsan, S. R., McFaline-Figueroa, J. L., Ramani, V., Saunders, L., Cao, J., Packer, J., Pliner, H. A., Jackson, D. L., Daza, R. M., Christiansen, L., et al. (2020) · Science · doi:10.1126/science.aax6234 · Cited by 401
Exploring single-cell data with deep multitasking neural networks
Amodio, M., van Dijk, D., Srinivasan, K., Chen, W. S., Mohsen, H., Moon, K. R., Campbell, A., Zhao, Y., Wang, X., Venkataswamy, M., et al. (2019) · Nature Methods · doi:10.1038/s41592-019-0576-7 · Cited by 297
Research works printed but not linked: 6
- Arjovsky, 2017, Proceedings of the 34th International Conference on Machine Learning, Volume 70 of Proceedings of Machine Learning Research, 214
- Dziugaite, 2015, Proceedings of the Thirty-First Conference on Uncertainty in Artificial Intelligence, UAI’15, 258
- Johansson, 2016, International Conference on Machine Learning, 3020
- Lopez, 2018, Advances in Neural Information Processing Systems, 6114
- Louizos, 2017, Advances in Neural Information Processing Systems, 6446
- Sohn, 2015, Learning structured output representation using deep conditional generative models, Advances in Neural Information Processing Systems, Vol. 28, 3483
Cited by
Benchmarking atlas-level data integration in single-cell genomics
Luecken, M. D., Büttner, M., Chaichoompu, K., Danese, A., Interlandi, M., Mueller, M. F., Strobl, D. C., Zappia, L., Dugas, M., Colomé-Tatché, M., et al. (2022) · Nature Methods · doi:10.1038/s41592-021-01336-8 · Cited by 1,438
Self-citationMapping single-cell data to reference atlases by transfer learning
Lotfollahi, M., Naghipourfar, M., Luecken, M. D., Khajavi, M., Büttner, M., Wagenstetter, M., Avsec, Ž., Gayoso, A., Yosef, N., Interlandi, M., et al. (2022) · Nature Biotechnology · doi:10.1038/s41587-021-01001-7 · Cited by 602
Self-citationJoint probabilistic modeling of single-cell multi-omic data with totalVI
Gayoso, A., Steier, Z., Lopez, R., Regier, J., Nazor, K. L., Streets, A., & Yosef, N. (2021) · Nature Methods · doi:10.1038/s41592-020-01050-x · Cited by 592
Predicting cellular responses to complex perturbations in high‐throughput screens
Lotfollahi, M., Klimovskaia Susmelj, A., De Donno, C., Hetzel, L., Ji, Y., Ibarra, I. L., Srivatsan, S. R., Naghipourfar, M., Daza, R. M., Martin, B., et al. (2023) · Molecular Systems Biology · doi:10.15252/msb.202211517 · Cited by 307
“Aside from CPA, existing methods (Lotfollahi et al , 2020; Russ-kikh et al , 2020) such as scGen (Lotfollahi et al , 2019) have also been shown capable of predicting single-cell perturbation responses when the dataset contains no combinatorial treatment or dose-dependent perturbations.”
BackgroundMethodSelf-citationEfficient and precise single-cell reference atlas mapping with Symphony
Kang, J. B., Nathan, A., Weinand, K., Zhang, F., Millard, N., Rumker, L., Moody, D. B., Korsunsky, I., & Raychaudhuri, S. (2021) · Nature Communications · doi:10.1038/s41467-021-25957-x · Cited by 231
scPerturb: harmonized single-cell perturbation data
Peidli, S., Green, T. D., Shen, C., Gross, T., Min, J., Garda, S., Yuan, B., Schumacher, L. J., Taylor-King, J. P., Marks, D. S., et al. (2024) · Nature Methods · doi:10.1038/s41592-023-02144-y · Cited by 194
“Multiple distance measures for scRNA-seq have been explored by the single-cell community in recent years, including Wasserstein distance computed in an optimal transport framework 22,23 , maximum mean discrepancy 24 , neighborhood-based measures 17,18 , and energy distance (E-distance) 20 .”
Deep generative modeling of transcriptional dynamics for RNA velocity analysis in single cells
Gayoso, A., Weiler, P., Lotfollahi, M., Klein, D., Hong, J., Streets, A., Theis, F. J., & Yosef, N. (2023) · Nature Methods · doi:10.1038/s41592-023-01994-w · Cited by 162
Self-citationMachine learning for perturbational single-cell omics
Ji, Y., Lotfollahi, M., Wolf, F. A., & Theis, F. J. (2021) · Cell Systems · doi:10.1016/j.cels.2021.05.016 · Cited by 137
Self-citationPopulation-level integration of single-cell datasets enables multi-scale analysis across samples
Donno, C. D., Hediyeh-zadeh, S., Moinfar, A. A., Wagenstetter, M., Zappia, L., Lotfollahi, M., & Theis, F. J. (2023) · Nature Methods · doi:10.1038/s41592-023-02035-2 · Cited by 152
“scPoli, therefore, 93 introduces two main modifications to the standard conditional variational autoencoder (CVAE) 94 architecture which constitutes the backbone of many successful methods addressing problems such 95 as data integration [5, 15, 24] and perturbation modeling [16, 29] in single-cell genomics.”
BackgroundSelf-citationAkbar, R., Bashour, H., Rawat, P., Robert, P. A., Smorodina, E., Cotet, T.-S., Flem-Karlsen, K., Frank, R., Mehta, B. B., Vu, M. H., et al. (2022) · mAbs · doi:10.1080/19420862.2021.2008790 · Cited by 113
“A more sophisticated strategy may include conditioning the model in such a way that the output is biased toward out-ofdistribution samples.(359)”
BackgroundBiologically informed deep learning to query gene programs in single-cell atlases
Lotfollahi, M., Rybakov, S., Hrovatin, K., Hediyeh-zadeh, S., Talavera-López, C., Misharin, A. V., & Theis, F. J. (2023) · Nature Cell Biology · doi:10.1038/s41556-022-01072-x · Cited by 100
Self-citationApplication of Deep Learning on Single-Cell RNA Sequencing Data Analysis: A Review
Brendel, M., Su, C., Bai, Z., Zhang, H., Elemento, O., & Wang, F. (2022) · Genomics Proteomics & Bioinformatics · doi:10.1016/j.gpb.2022.11.011 · Cited by 122
“com/theislab/trVAE Using conditional VAE with MMD regularization in latent space 2020 [102] scDGN DFNN https://github.”
BackgroundMethodDigital twins in healthcare: a comprehensive review and future directions
Rudsari, H. K., Tseng, B., Zhu, H., Song, L., Gu, C., Roy, A., Irajizad, E., Butner, J. D., Long, J. P., & Do, K.-A. (2025) · Frontiers in Digital Health · doi:10.3389/fdgth.2025.1633539 · Cited by 146
The future of rapid and automated single-cell data analysis using reference mapping
Lotfollahi, M., Hao, Y., Theis, F. J., & Satija, R. (2024) · Cell · doi:10.1016/j.cell.2024.03.009 · Cited by 86
Self-citationHe, S., Jin, Y., Nazaret, A., Shi, L., Chen, X., Rampersaud, S., Dhillon, B. S., Valdez, I. K., Friend, L. E., Fan, J. L., et al. (2024) · Nature Biotechnology · doi:10.1038/s41587-024-02173-8 · Cited by 100
“Deep generative models parameterized by neural networks have proven effective in analyzing single-cell RNA expression data (scvi-tools 19 , scVI 20 , totalVI 21 , scArches 22 , trVAE 23 , scANVI 24 , MrVI 25 and so on).”
Optimal transport for single-cell and spatial omics
Bunne, C., Schiebinger, G., Krause, A., Regev, A., & Cuturi, M. (2024) · Nature Reviews Methods Primers · doi:10.1038/s43586-024-00334-2 · Cited by 71
Integration of Single-Cell RNA-Seq Datasets: A Review of Computational Methods
Ryu, Y., Han, G. H., Jung, E., & Hwang, D. (2023) · Molecules and Cells · doi:10.14348/molcells.2023.0009 · Cited by 80
“To address these issues, several methods using a generative model with variational autoencoder (e.g., scVI [Lopez et al., 2018], scGen [Lotfollahi et al., 2019], and trVAE [Lotfollahi et al., 2020]) have been developed.”
Contrastive learning enables rapid mapping to multimodal single-cell atlas of multimillion scale
Yang, M., Yang, Y., Xie, C., Ni, M., Liu, J., Yang, H., Mu, F., & Wang, J. (2022) · Nature Machine Intelligence · doi:10.1038/s42256-022-00518-z · Cited by 80
“13–17 and Supplementary Tables 4–6 and 8.”
MethodIntegration and Querying of Multimodal Single-Cell Data with PoE-VAE
Litinetskaya, A., Shulman, M., Curion, F., Szałata, A., Omidi, A., Lotfollahi, M., & Theis, F. J. (2025) · Lecture notes in computer science · doi:10.1007/978-3-031-90252-9_36 · Cited by 66
Self-citationDeep learning in spatially resolved transcriptomics: a comprehensive technical view
Zahedi, R., Ghamsari, M. R. E., Argha, A., MacPhillamy, C., Beheshti, A., Alizadehsani, R., Lovell, N. H., Lotfollahi, M., & Alinejad‐Rokny, H. (2024) · Briefings in Bioinformatics · doi:10.1093/bib/bbae082 · Cited by 94
“Results: StLearn examined twelve human and mouse brain datasets and achieved a greater Adjusted Rand Index (ARI) value compared to the SpatialLIBD (a graph-based clustering method) [4] and detected two more tissue layers than Seurat.”
Self-citationA mini-review on perturbation modelling across single-cell omic modalities
Gavriilidis, G. I., Vasileiou, V., Orfanou, A., Ishaque, N., & Psomopoulos, F. (2024) · Computational and Structural Biotechnology Journal · doi:10.1016/j.csbj.2024.04.058 · Cited by 54
“An alternative model is trVAE [53], which incorporates maximum mean discrepancy (MMD) in a conditional variational autoencoder (CVAE) framework.”
MethodLearning interpretable cellular responses to complex perturbations in high-throughput screens
Lotfollahi, M., Susmelj, A. K., De Donno, C., Ji, Y., Ibarra, I. L., Wolf, F. A., Yakubova, N., Theis, F. J., & Lopez-Paz, D. (2021) · bioRxiv (Cold Spring Harbor Laboratory) · doi:10.1101/2021.04.14.439903 · Cited by 48
“Recently, DL methods have been used to model gene 47 expression latent spaces from scRNA-seq data [22–25], and describe and predict single-cell responses 48 [16, 17, 20, 26].”
BackgroundMethodInfluentialSelf-citationMachine learning for single-cell genomics data analysis
Raimundo, F., Meng-Papaxanthos, L., Vallot, C., & Vert, J. (2021) · Current Opinion in Systems Biology · doi:10.1016/j.coisb.2021.04.006 · Cited by 51
Toseef, M., Petinrin, O. O., Wang, F., Rahaman, S., Liu, Z., Li, X., & Wong, K. (2023) · Briefings in Bioinformatics · doi:10.1093/bib/bbad254 · Cited by 47
Benchmarking algorithms for generalizable single-cell perturbation response prediction
Wei, Z., Wang, Y., Gao, Y., Wang, S., Li, P., Si, D., Gao, Y., Wu, S., Li, D., Dong, K., et al. (2026) · Nature Methods · doi:10.1038/s41592-025-02980-0 · Cited by 42
RxRx1: A Dataset for Evaluating Experimental Batch Correction Methods
Sypetkowski, M., Rezanejad, M., Saberian, S., Kraus, O. Z., Urbanik, J., Taylor, J., Mabey, B., Victors, M. L., Yosinski, J., Sereshkeh, A. R., et al. (2023) · IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops (CVPRW) · doi:10.1109/cvprw59228.2023.00451 · Cited by 75
A Roadmap for a Consensus Human Skin Cell Atlas and Single-Cell Data Standardization
Almet, A. A., Yuan, H., Annusver, K., Ramos, R., Liu, Y., Wiedemann, J., Sorkin, D. H., Xu, N., Sonkoly, E., Haniffa, M. A., et al. (2023) · Journal of Investigative Dermatology · doi:10.1016/j.jid.2023.03.1679 · Cited by 34
Building, benchmarking, and exploring perturbative maps of transcriptional and morphological data
Çelik, S., Hütter, J.-C., Carlos, S. M., Lazar, N. H., Mohan, R., Tillinghast, C. A., Biancalani, T., Fay, M. M., Earnshaw, B. A., & Haque, I. S. (2024) · PLoS Computational Biology · doi:10.1371/journal.pcbi.1012463 · Cited by 48
Kopp, W., Akalin, A., & Ohler, U. (2022) · Nature Machine Intelligence · doi:10.1038/s42256-022-00443-1 · Cited by 31
PerturbNet predicts single-cell responses to unseen chemical and genetic perturbations
Yu, H., Qian, W., Song, Y., & Welch, J. D. (2025) · Molecular Systems Biology · doi:10.1038/s44320-025-00131-3 · Cited by 57
“In 2020, Lotfollahi et al introduced a conditional variational autoencoder framework, balancing representations under two treatment conditions using a similarity score based on counter-factual inference (Johansson et al, 2016; Lotfollahi et al, 2020).”
Jiang, Q., Chen, S., Chen, X., & Jiang, R. (2024) · Bioinformatics · doi:10.1093/bioinformatics/btae265 · Cited by 29
“Representative methods of this approach include scGen (Lotfollahi et al. 2019), trVAE (Lotfollahi et al. 2020), and scVIDR (Kana et al. 2023). scGen employs a variational autoencoder (VAE) (Kingma and Welling 2014) to capture latent encodings and employs a vector algorithm to calculate the…”
MethodJoint trajectory inference for single-cell genomics using deep learning with a mixture prior
Du, J., Chen, T., Gao, M., & Wang, J. (2024) · Proceedings of the National Academy of Sciences · doi:10.1073/pnas.2316256121 · Cited by 42
“To eliminate all differences between the two groups, the MMD loss is characterized in the latent space by the following formulation: where k ( · , · ) is the multi-scale RBF kernel proposed by [38].”
BackgroundA Systematic Comparison of Single-Cell Perturbation Response Prediction Models
Li, L., You, Y., Fu, Y., Liao, W., Fan, X., Lu, S., Cao, Y., Li, B., Ren, W., Kong, J., et al. (2024) · bioRxiv (Cold Spring Harbor Laboratory) · doi:10.1101/2024.12.23.630036 · Cited by 25
PerturbNet predicts single-cell responses to unseen chemical and genetic perturbations
Yu, H., & Welch, J. D. (2022) · doi:10.1101/2022.07.20.500854 · Cited by 57
“(2020) proposed a conditional variational autoencoder (VAE) framework with representa62 tions under two treatment conditions [13] balanced using a similarity score of their counterfactual 63 inference [14].”
MethodIntegrating single-cell RNA-seq datasets with substantial batch effectsPreprint
Hrovatin, K., Moinfar, A. A., Zappia, L., Tejada-Lapuerta, A., Lengerich, B. J., Kellis, M., & Theis, F. J. (2023) · bioRxiv (Cold Spring Harbor Laboratory) · doi:10.1101/2023.11.03.565463 · Cited by 24
“They work by adding batch information as a covariate to the encoder and decoder, thus enabling learning of a batch-free latent space 20,21 .”
Self-citationHausmann, F., Ergen, C., Khatri, R., Marouf, M., Hänzelmann, S., Gagliani, N., Huber, S., Machart, P., & Bonn, S. (2023) · Genome biology · doi:10.1186/s13059-023-03049-x · Cited by 23
Weakly supervised learning uncovers phenotypic signatures in single-cell dataPreprint
Litinetskaya, A., Hediyeh-zadeh, S., Moinfar, A. A., Lotfollahi, M., & Theis, F. J. (2024) · bioRxiv (Cold Spring Harbor Laboratory) · doi:10.1101/2024.07.29.605625 · Cited by 20
Self-citationGenePert: Leveraging GenePT Embeddings for Gene Perturbation PredictionPreprint
Chen, Y. T., & Zou, J. (2024) · bioRxiv (Cold Spring Harbor Laboratory) · doi:10.1101/2024.10.27.620513 · Cited by 31
“While a few computational methods perturbation effects on a single-cell resolution [Lotfollahi et al., 2020, Dong et al., 2023, Bunne et al., 2023], most prediction and benchmarking efforts have focused on estimating the average perturbation effect, defined as ¯ X p − ¯ X c .”
BackgroundMethodResultInfluentialKim, J., & Seok, J. (2024) · Briefings in Bioinformatics · doi:10.1093/bib/bbae325 · Cited by 18
“For instance, Lotfollahi et al. [38] presented trVAE, and Russkikh et al. [39] proposed stVAE, both of which implement style-transfer methodologies in conditional VAE (CVAE) to generate gene expression. trVAE overcomes the limitations of CVAE using the maximum mean discrepancy in the decoder layer.”
MethodTransfer learning of clinical outcomes from preclinical molecular data, principles and perspectives
Kowald, A., Barrantes, I., Möller, S., Palmer, D. H., Escobar, H. M., Schwerk, A. M. K., & Fuellen, G. (2022) · Briefings in Bioinformatics · doi:10.1093/bib/bbac133 · Cited by 19
Polyphony: an Interactive Transfer Learning Framework for Single-Cell Data Analysis
Cheng, F., Keller, M. S., Qu, H., Gehlenborg, N., & Wang, Q. (2022) · IEEE Transactions on Visualization and Computer Graphics · doi:10.1109/tvcg.2022.3209408 · Cited by 24
Active Learning for Data Quality Control: A Survey
Li, N., Qi, Y., Li, C., & Zhao, Z. (2024) · Journal of Data and Information Quality · doi:10.1145/3663369 · Cited by 26
“Addressing the representation disentanglement by exploring intrinsic supervision signals Lotfollahi et al. [110] trVAE Image and RNA-seq data More accurate OOD generation compared with standard CVAE implementations.”
Moinfar, A. A., & Theis, F. J. (2024) · bioRxiv (Cold Spring Harbor Laboratory) · doi:10.1101/2024.11.06.622266 · Cited by 24
Self-citationIntegrating single-cell RNA-seq datasets with substantial batch effects
Hrovatin, K., Moinfar, A. A., Zappia, L., Parikh, S., Tejada-Lapuerta, A., Lengerich, B. J., Kellis, M., & Theis, F. J. (2025) · BMC Genomics · doi:10.1186/s12864-025-12126-3 · Cited by 51
“Multiple approaches have been proposed for promoting indistinguishability of latent distributions across batches, such as maximum mean discrepancy [31, 69], contrastive mixture of posteriors misalignment penalty [32], and disentanglement of batch and biology-related latent components [33].”
Self-citationMachine learning to dissect perturbations in complex cellular systems
Monfort‐Lanzas, P., Rungger, K., Madersbacher, L., & Hackl, H. (2025) · Computational and Structural Biotechnology Journal · doi:10.1016/j.csbj.2025.02.028 · Cited by 23
“Meanwhile, transfer variational autoencoder extends the VAE framework to predict single-cell perturbation responses while addressing batch effects and domain adaptation challenges [29].”
Forecasting cellular states: from descriptive to predictive biology via single-cell multiomics
Stein-O’Brien, G., Ainslie, M. C., & Fertig, E. J. (2021) · Current Opinion in Systems Biology · doi:10.1016/j.coisb.2021.03.008 · Cited by 15
Deep learning models will shape the future of stem cell research
Ouyang, J. F., Chothani, S. P., & Rackham, O. J. L. (2023) · Stem Cell Reports · doi:10.1016/j.stemcr.2022.11.007 · Cited by 14
Application of Deep Learning for Single Cell Multi-Omics: A State-of-the-Art Review
Wani, S. A., Khan, S. A., & Quadri, S. (2025) · Archives of Computational Methods in Engineering · doi:10.1007/s11831-025-10230-x · Cited by 15
“The trVAE [183] employs VAE to predict perturbation response under multiple biological conditions.”
Interpretation, extrapolation and perturbation of single cells
Dimitrov, D., Schrod, S., Rohbeck, M., & Stegle, O. (2026) · Nature Reviews Genetics · doi:10.1038/s41576-025-00920-4 · Cited by 37
Fritz, D., Inamo, J., & Zhang, F. (2023) · Frontiers in Immunology · doi:10.3389/fimmu.2022.1076700 · Cited by 13
Zhang, S., Fang, X., Chang, M., Zheng, M., Guo, L., Xu, Y., Shu, J., Nie, Q., & Li, Z. (2024) · BMC Genomics · doi:10.1186/s12864-024-11030-6 · Cited by 14
Wan, H. S., Chen, L., & Deng, M. (2022) · Genomics Proteomics & Bioinformatics · doi:10.1016/j.gpb.2022.12.008 · Cited by 11
“It can be compatible with various autoencoder models such as scANVI [12] and trVAE [13].”
MethodscJVAE: A novel method for integrative analysis of multimodal single-cell data
Wani, S. A., Khan, S. A., & Quadri, S. M. K. (2023) · Computers in Biology and Medicine · doi:10.1016/j.compbiomed.2023.106865 · Cited by 10
Joint Trajectory Inference for Single-cell Genomics Using Deep Learning with a Mixture PriorPreprint
Du, J., Chen, T., Gao, M., & Wang, J. (2020) · bioRxiv (Cold Spring Harbor Laboratory) · doi:10.1101/2020.12.26.424452 · Cited by 8
Guo, W., Li, X., Wang, D., Yan, N., Hu, Q., Yang, F., Zhang, X., Yao, J., & Gu, J. (2024) · Genome biology · doi:10.1186/s13059-024-03436-y · Cited by 8
“Aligning at cluster-level, through direct integrated clustering [6, 7] or nonlinear distribution comparison [8–11], has emerged as a widely utilized strategy and leads to some valuable biological discoveries.”
SeuratIntegrate: an R package to facilitate the use of integration methods with Seurat
Specque, F., Barré, A., Nikolski, M., & Chalopin, D. (2025) · Bioinformatics · doi:10.1093/bioinformatics/btaf358 · Cited by 9
Out-of-distribution prediction with disentangled representations for single-cell RNA sequencing dataPreprint
Lotfollahi, M., Dony, L., Agarwala, H., & Theis, F. J. (2021) · bioRxiv (Cold Spring Harbor Laboratory) · doi:10.1101/2021.09.01.458535 · Cited by 6
Self-citationStarfysh reveals heterogeneous spatial dynamics in the breast tumor microenvironmentPreprint
He, S., Jin, Y., Nazaret, A., Shi, L., Chen, X., Rampersaud, S., Dhillon, B. S., Valdez, I. K., Friend, L. E., Fan, J. L., et al. (2022) · bioRxiv (Cold Spring Harbor Laboratory) · doi:10.1101/2022.11.21.517420 · Cited by 7
Wu, Y., Liu, J., Liu, S., Xiao, Y., Zhang, S., & Li, L. (2025) · Briefings in Bioinformatics · doi:10.1093/bib/bbaf126 · Cited by 12
“VAE-based generative models have been more popularly used [21, 22].”
Computational modeling of single-cell dynamics data
Guo, W., Chen, Z., & Gu, J. (2025) · Briefings in Bioinformatics · doi:10.1093/bib/bbaf305 · Cited by 10
scCausalVI disentangles single-cell perturbation responses with causality-aware generative model
An, S., Cho, J.-W., Cao, K., Xiong, J., Hemberg, M., & Wan, L. (2025) · Cell Systems · doi:10.1016/j.cels.2025.101443 · Cited by 9
Friedrich, V. D., Pennitz, P., Wyler, E., Adler, J. M., Postmus, D., Müller, K., Alves, L. G. T., Prigann, J., Pott, F., Vladimirova, D., et al. (2024) · EBioMedicine · doi:10.1016/j.ebiom.2024.105312 · Cited by 8
“Thus, VAEs have been utilised in scRNAseq data analyses to model sparse data, 27 integrate data, 28,29 generate data 29 – 31 and perform reference mapping.”
Toward informed batch correction for single-cell transcriptome integration
Li, S., Luecken, M. D., Marioni, J. C., Teichmann, S. A., & He, P. (2026) · Nature Computational Science · doi:10.1038/s43588-025-00943-1 · Cited by 18
Walbech, J. S., Kinalis, S., Winther, O., Nielsen, F. C., & Bagger, F. O. (2021) · Cells · doi:10.3390/cells11010085 · Cited by 6
Modeling the cell-type-specific mesoscale murine connectome with anterograde tracing experiments
Koelle, S., Mastrovito, D., Whitesell, J. D., Hirokawa, K. E., Zeng, H., Meilă, M., Harris, J. A., & Mihalaş, Ş. (2023) · Network Neuroscience · doi:10.1162/netn_a_00337 · Cited by 5
Integration of single‐cell proteomic datasets through distinctive proteins in cell clusters
Koca, M. B., & Sevilgen, F. E. (2023) · PROTEOMICS · doi:10.1002/pmic.202300282 · Cited by 5
scEMB: Learning context representation of genes based on large-scale single-cell transcriptomicsPreprint
Hsieh, K., Chu, Y., Li, X., Pilie, P. G., & Dai, Y. (2024) · bioRxiv (Cold Spring Harbor Laboratory) · doi:10.1101/2024.09.24.614685 · Cited by 8
“Moreover, these embedding-based methods 11 [Autoencoder 12 , VAE 13–16 , GNN 17 ], has been widely used to along with the cellular perturbation tasks along with single-cell CRISPR data, to compress the transcriptomics data (mainly single-cell) into the latent space and compare the effects of…”
MethodSuhendar, H., Iswanto, B. H., Pratama, M. R., Silambi, M. S., Trıyana, K., & Santoso, I. (2025) · Applied Food Research · doi:10.1016/j.afres.2025.101445 · Cited by 4
Biochemical-knowledge-driven machine learning pipeline for generating potent antimicrobial peptides
Yang, D., Li, Y., Li, C., Zhang, Q., HUANG, J., Li, X., & Gao, P. (2026) · Briefings in Bioinformatics · doi:10.1093/bib/bbag115 · Cited by 4
“The model is trained to minimize a composite loss function comprising three terms: reconstruction loss, Kullback–Leibler (KL) divergence and Jacobian disentanglement regularization [11, 37].”
Conditional Monge Gap enables generalizable single-cell perturbation modelling
Driessen, A., Rajwade, D. A., Harsanyi, B., Rapsomaniki, M., & Born, J. (2026) · Nature Machine Intelligence · doi:10.1038/s42256-026-01242-8 · Cited by 16
“Previous work on single-cell perturbation modelling includes the landmark autoencoder method scGen 8 , which has been extended to capture combinatorial perturbations 9–11 and demonstrates that latent variable models can learn perturbation-associated shifts from limited observations.”
McLean, A. K., Reynolds, G., & Pratt, A. G. (2024) · Biomedicines · doi:10.3390/biomedicines12061297 · Cited by 4
Piochi, L. F., Gaspar, A. T., Rosário‐Ferreira, N., Preto, A. J., & Moreira, I. S. (2022) · Advances in protein chemistry and structural biology · doi:10.1016/bs.apcsb.2022.05.006 · Cited by 3
Evaluation of Computational Methods for Single Cell Multi-Omics Integration
Wani, S. A., & Quadri, S. M. K. (2023) · Procedia Computer Science · doi:10.1016/j.procs.2023.01.246 · Cited by 5
GNODEVAE: a graph-based ODE-VAE enhances clustering for single-cell data
Fu, Z., Chen, C., Wang, S., Wang, J., & Chen, S. (2025) · BMC Genomics · doi:10.1186/s12864-025-11946-7 · Cited by 3
“…demonstrate VAE’s utility: scVI [21] models gene expression with zero-inflated negative binomial distributions while handling batch effects; trVAE predicts cellular responses to perturbations [40]; and MM-VAE integrates multi-modal data for comprehensive cell state understanding [41].”
An expanded role for single-cell chemical genomics profiling in drug discovery
Wyatt, A., Hoffer-Hawlik, K., Giglio, R. M., Azizi, E., & McFaline‐Figueroa, J. L. (2026) · Biochemical Journal · doi:10.1042/bcj20253273 · Cited by 4
“…trVAE model that uses maximum mean discrepancy regularization to force cross-condition similarities and enable out-of-distribution predictions [129]. scVIDR additionally uses log-linear interpolation to predict cell type-specific, dose-dependent gene expression changes across unseen doses…”
Roseth, I. A., Hatscher, L., Schiller, C., Johnstuen, H. S. S., Slørdahl, T. S., Hov, H., Schapiro, D., & Standal, T. (2026) · Blood Cancer Discovery · doi:10.1158/2643-3230.bcd-25-0334 · Cited by 8
Fan, Y., Su, Y., Hao, G., Wang, F., Chen, X., Wong, K., & Li, X. (2026) · Proceedings of the National Academy of Sciences · doi:10.1073/pnas.2519870123 · Cited by 3
Wang, Y., Lu, T., Chen, X., Yao, Z., & Wong, K. (2025) · Bioengineering · doi:10.3390/bioengineering12080884 · Cited by 4
“Representative methods using the variational autoencoder (VAE) structure include scGen [12], trVAE [16], and scVIDR [17].”
Kion-Crosby, W., & Barquist, L. (2024) · Bioinformatics Advances · doi:10.1093/bioadv/vbae066 · Cited by 3
“A number of studies have explored architecture choices in VAEs, but only by examination of the test score from held-out data on eukaryotic data sets (Lotfollahi et al. 2020, 2023, Lotfollahi et al. , 2021, Chow et al. 2022).”
BackgroundLyu, Y., Lin, S. H., Wu, H., & Li, Z. (2024) · Bioinformatics · doi:10.1093/bioinformatics/btae537 · Cited by 3
“Notable deep learning approaches include scVI (Lopez et al. 2018), trVAE (Lotfollahi et al. 2020), and scGen (Lotfollahi et al. 2019).”
MethodPatient Informed Domain Adaptation Improves Clinical Drug Response PredictionPreprint
Anastopoulos, I. N., Seninge, L., Ding, H., & Stuart, J. M. (2021) · bioRxiv (Cold Spring Harbor Laboratory) · doi:10.1101/2021.08.07.455527 · Cited by 5
Zhu, H., Asiaee, A., Azinfar, L., Li, J., Liang, H., Irajizad, E., Do, K.-A., & Long, J. P. (2025) · Briefings in Bioinformatics · doi:10.1093/bib/bbaf426 · Cited by 9
“Several studies, such as [15, 16] and [17], introduced variational autoencoder-based methods to predict out-of-sample single-cell perturbation responses.”
What Do Single-Cell Models Already Know About Perturbations?
Bjerregaard, A., Prada-Luengo, I., Das, V., & Krogh, A. (2025) · Genes · doi:10.3390/genes16121439 · Cited by 4
“Recent methods explicitly learn perturbation mappings with supervision or mechanistic priors, including scGen and trVAE [4,5], regulatory network-based approaches such as CellOracle [6], and optimal transport formulations [7].”
Bio-Inspired Generative Network with Knowledge Integration
Batbaatar, E., & Ryu, K. H. (2025) · Applied Sciences · doi:10.3390/app152412918 · Cited by 2
Dashtgoli, D. S., Narsilio, G. A., Alqawasmeh, Q. I., Giustiniani, M., Busetti, M., & Cherubini, C. (2026) · Journal of Rock Mechanics and Geotechnical Engineering · doi:10.1016/j.jrmge.2026.03.025 · Cited by 3
Wang, L., Nie, R., Zhang, Z., Gu, W., Wang, S., Wang, A., Zhang, J., & Cai, J. (2023) · Cell Reports Methods · doi:10.1016/j.crmeth.2023.100558 · Cited by 2
“To predict the effect of unseen covariate combinations on transcription profile based on monocovariate and existing covariate combinations is a kind of conditional OOD generation problem.(11,12) InClust can provide a supervised solution for this problem.”
MethodFriedrich, V. D., Pennitz, P., Wyler, E., Adler, J. M., Postmus, D., Alves, L. G. T., Prigann, J., Pott, F., Vladimirova, D., Hoefler, T., et al. (2024) · bioRxiv (Cold Spring Harbor Laboratory) · doi:10.1101/2024.01.11.574849 · Cited by 2
Unify: Learning Cellular Evolution with Universal Multimodal EmbeddingsPreprint
Zhong, H., Han, W., Cui, G., Gómez-Cabrero, D., Tegner, J. N., Gao, X., & Aranda, M. (2025) · bioRxiv (Cold Spring Harbor Laboratory) · doi:10.1101/2025.09.07.674681 · Cited by 2
“Additionally, existing integration approaches typically produce either corrected expression profiles restricted to one-to-one orthologous genes 14-17 or integrated embeddings 6, 10, 18-20 that lack direct interpretability.”
A systematic comparison of single-cell perturbation response prediction models
Li, L., You, Y., Fu, Y., Liao, W., Fan, X., Lu, S., Cao, Y., Li, B., Ren, W., Kong, J., et al. (2026) · Science Advances · doi:10.1126/sciadv.aed3414 · Cited by 1
Inference of differentiation trajectories by transfer learning across biological processes
Jumde, G., Spanjaard, B., & Junker, J. P. (2023) · Cell Systems · doi:10.1016/j.cels.2023.12.002 · Cited by 2
Annotation of single-cell clusters using marker genes within and across species
Kundu, S., Chau, T., Maroof, D. S., & Li, S. (2025) · Elsevier eBooks · doi:10.1016/b978-0-443-23736-2.00015-0 · Cited by 1
Zhou, W., Fan, X., Li, L., Zheng, J., Liu, X., Jin, W., & Tian, L. (2026) · Briefings in Bioinformatics · doi:10.1093/bib/bbag432 · Cited by 3
“Additionally, to mitigate batch effects, Garfield applies MMD [ 40] loss as an alignment strategy, which is widely utilized in batch correction [ 32, 41].”
scVDM: A Diffusion Model Integrated With Conditional VAE for Generative Single-Cell Tasks
Peng, D., Xie, L., Yang, H. T., & Zhang, Y. (2025) · IEEE Journal of Biomedical and Health Informatics · doi:10.1109/jbhi.2025.3585960 · Cited by 1
“Furthermore, Lotfollahi et al. [23] also proposed a VAE-based model called trVAE in 2020.”
Luo, Y., Liu, L., Fu, Y., Deng, Y. X., & Tang, L. (2025) · Journal of Biomedical Informatics · doi:10.1016/j.jbi.2025.104915 · Cited by 1
Identifying an optimal perturbation to induce a desired cell state by generative deep learning
Han, Y.-H., Kim, H., Lee, C., & Cho, K. (2025) · Cell Systems · doi:10.1016/j.cels.2025.101405 · Cited by 2
scREPA: Predicting single-cell perturbation responses with cycle-consistent representation alignment
Wang, Y., Chen, X., Li, X., & Wong, K. (2025) · Computational Biology and Chemistry · doi:10.1016/j.compbiolchem.2025.108712 · Cited by 5
Integration of unpaired and heterogeneous clinical flow cytometry data
Phuycharoen, M., Kästele, V., Williams, T., Lijing Lin, A., Hussell, T., Grainger, J. R., & Rattray, M. (2026) · iScience · doi:10.1016/j.isci.2026.114937 · Cited by 1
“10,11 VAE-based normalization can be performed with latent arithmetic 12 or conditional au-toencoding, 13 although the latter may still capture batch-corre-lated information, which can be mitigated with additional loss terms like MMD. 10 A major challenge in all integration tasks is that unknown…”
CFM-GP: unified conditional flow matching to learn gene perturbation across cell types
Abir, A. R., Dip, S. A., & Choi, J. M. (2026) · NAR Genomics and Bioinformatics · doi:10.1093/nargab/lqag087 · Cited by 5
“Early approaches such as scGen [ 17 ], CVAE [ 20 ], and trVAE [ 18 ] use latent space arithmetic to model expression shifts from control to perturbed states.”
Yeh, C., Chen, Z.-G., Liou, C.-Y., & Chen, M. (2023) · Bioengineering · doi:10.3390/bioengineering10090996 · Cited by 3
TextSMatch: Safe Semi-supervised Text Classification with Domain Adaption
Xu, Y., Lin, G., Zeng, N., Qu, Y., & Zeng, K. (2022) · Communications in computer and information science · doi:10.1007/978-981-19-6142-7_33 · Cited by 1
Metadata
- Authors
- Lotfollahi, M., Naghipourfar, M., Theis, F. J., & Wolf, F. A.
- Affiliations
- Helmholtz Munich (DE)
- Corresponding author
- Theis, Fabian J
- Wolf, F Alexander
- Venue
- Bioinformatics
- Publisher
- Oxford University Press (OUP)
- Published
- 2020-12
- Volume
- 36
- Issue
- Supplement_2
- Pages
- i610-i617
- ISSN
- 1367-4803 (ISSN-L)
- 1367-4811
- OpenAlex
- W3117464799
- Cited by
- 130
- References
- 14
- Citations by year
- 2026: 25 · 2025: 31 · 2024: 24 · 2023: 25 · 2022: 12 · 2021: 12 · 2020: 1
- FWCI
- 4.05
- Type
- journal article
- Language
- en
- Open access
- Licence
- Locations
- Bioinformatics (publishedVersion)
- PubMed (publishedVersion)
- Site cant be reached (submittedVersion, other-oa)
- PuSH - Publication Server of Helmholtz Zentrum München (submittedVersion)
- Funders
- BMBF (01IS18036A, 01IS18053A)
- German Research Foundation (ZT-I-0007)
- Chan Zuckerberg Initiative DAF (182835)
- Joachim Herz Stiftung
- Topics
- Life Sciences › Biochemistry, Genetics and Molecular Biology › Molecular Biology › Single-cell and spatial transcriptomics
- Physical Sciences › Computer Science › Computer Vision and Pattern Recognition › Generative Adversarial Networks and Image Synthesis
- Physical Sciences › Computer Science › Artificial Intelligence › Gaussian Processes and Bayesian Inference
- Keywords
- conditional variational autoencoder, maximum mean discrepancy, style transfer, unpaired data, single-cell RNA sequencing, cell type-specific gene expression, perturbation response prediction, minority class, generative models
- MeSH terms
- Exome Sequencing, RNA-Seq, Reproducibility of Results
About these checks
Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.
Statistics and arithmetic are recomputed, not judged.
Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.
What counts as a claim ›How we measure confidence ›How we check this ›
Are you an author, or do you see a problem with this page? Request a review
Data sources
Bibliographic data: Semantic Scholar, OpenAlex, Crossref.
Contains information from the Semantic Scholar Academic Graph Datasets, made available under the ODC Attribution License (ODC-BY 1.0).
Semantic Scholar’s data is described in: Kinney, R., et al. (2023). The Semantic Scholar Open Data Platform. arXiv:2301.10140.