Table 4 Comparisons of computational strategies for spatial transcriptomics data analysis
From: Advances in spatial transcriptomics and related data analysis strategies
Package name | Year | Journal | Developer | Algorithm | Programming language | Usage | Limitation | References |
|---|---|---|---|---|---|---|---|---|
Seurat | 2015 | Nat Biotechnol | Satija, R. et al. | L1-constrained linear model | R | Clusters identification, data integration, gene imputation | Suitable for only certain platforms of ST | [22] |
SpatialDE | 2018 | Nat Methods | Svensson, V. et al. | Gaussian process regression | Python | Spatially-variable genes identification | Heavy computational burden | [55] |
trendsceek | 2018 | Nat Methods | Edsgärd, D. et al. | Marked point process | R | Spatially-variable genes identification | Heavy computational burden | [56] |
GCNG | 2020 | Genome Biol | Yuan, Y. and Bar-Joseph, Z | Graph convolutional network | Python | Cellular interaction | Needs to be optimized when performed on individual datasets | [63] |
SpaGE | 2020 | Nucleic Acids Res | Abdelaal, T. et al. | Domain adaptation model | Python | Data integration, gene imputation | Limited range of genes included in the model | [62] |
SpaOTsc | 2020 | Nat Commun | Cang, Z. and Nie, Q | Structured optimal transport model | Python | Cellular interaction | Ignores time delay in cellular communication | [64] |
SPARK | 2020 | Nat Methods | Sun, S. et al. | Generalized linear spatial model with penalized quasi-likelihood | R | Spatially-variable genes identification | Performs better for certain datasets but not all | [57] |
SpatialCPie | 2020 | BMC Bioinformatics | Bergenstråhle, J. et al. | N/A | R | Clusters identification | Limited usage | [87] |
stLearn | 2020 | bioRxiv | Pham, D. et al. | Transfer learning with a convolutional neural network, pseudo-space–time algorithm | Python | Clusters identification, cellular interaction, region annotation, spatial trajectories | Suitable for only certain platforms of ST | [69] |
stereoscope | 2020 | Commun Biol | Andersson, A. et al. | Negative binomial distribution with maximum a posteriori estimation | Python | Data integration, spatial decomposition | Needs more deeply sequenced data | [88] |
STUtility | 2020 | BMC Genomics | Bergenstråhle, J. et al. | Non-negative matrix factorization | R | Clusters identification, spatially-variable genes identification | Suitable for only certain platforms of ST | [52] |
SPATA | 2020 | bioRxiv | Kueckelhaus, J. et al. | Shared-nearest neighbor clustering, pattern recognition, Bayesian model | R | Spatial trajectories, spatial CNV identification | Suitable for only certain platforms of ST | [67] |
BayesSpace | 2021 | Nat Biotechnol | Zhao, E. et al. | Bayesian model with a Markov random field | R | Clusters identification | Suitable for only certain platforms of ST | [53] |
DSTG | 2021 | Brief Bioinform | Song, Q. and Su, J | Semi-supervised graph-based convolutional network | Python | Data integration, spatial decomposition | Black-box problem of the Artificial Intelligence model | [89] |
Giotto | 2021 | Genome Biol | Dries, R. et al. | A wide range of algorithms containing loess regression, HMRF, etc | R | Clusters identification, cellular interaction | Suitable for only certain platforms of ST | [90] |
SOMDE | 2021 | Bioinformatics | Hao, M. et al. | Gaussian process | Python | Spatially-variable genes identification | Loss of some spatial details | [91] |
MULTILAYER | 2021 | Cell Syst | Moehlin, J. et al. | Pattern recognition, community detection, agglomerative clustering | Python | Clusters identification, region annotation | May perform not as well on low-resolution data | [68] |
SpaGCN | 2021 | Nat Methods | Hu, J. et al. | Graph convolutional network | Python | Clusters identification, spatially-variable genes identification, region annotation | Potential disagreement between actual tissue structure and detected spatial domains | [54] |
SpatialDWLS | 2021 | Genome Biol | Dong, R. and Yuan, G.C | Weighted least squares | R | Data integration, spatial decomposition | Causes bias when removing some cell types | [61] |
SPOTlight | 2021 | Nucleic Acids Res | Elosua-Bayes, M. et al. | Seeded non-negative matrix factorization regression | R | Data integration, spatial decomposition | Does not consider the information of capturing position | [59] |
Tangram | 2021 | Nat Methods | Biancalani, T. et al. | Nonconvex optimization by a deep learning framework | Python | Data integration, spatial decomposition, gene imputation | Performs not as well on higher-density tissues | [23] |
CARD | 2022 | Nat Biotechnol | Ma, Y. and Zhou, X | Conditional autoregressive model with a non-negative matrix factorization model | R | Data integration, spatial decomposition | Does not incorporate histology image | [58] |
cell2location | 2022 | Nat Biotechnol | Kleshchevnikov, V. et al. | Bayesian model | Python | Data integration, spatial decomposition | Needs refinement for higher-resolution ST assays | [92] |
CellTrek | 2022 | Nat Biotechnol | Wei, R. et al. | Coembedding and metric learning | R | Data integration, spatial decomposition | Sparse maps of cells in certain regions of tissue | [70] |
RCTD | 2022 | Nat Biotechnol | Cable, D.M. et al. | Poisson distribution with maximum-likelihood estimation | R | Data integration, spatial decomposition | Disagreement of cell types between reference and spatial data | [60] |
STAGATE | 2022 | Nat Commun | Dong, K. and Zhang, S | Graph attention auto-encoder | Python | Clusters identification, spatially-variable genes identification, gene imputation | Does not integrate histology images well | [93] |
SpatialInferCNV | 2022 | Nature | Erickson, A. et al. | Hidden markov model | R | Spatial CNV identification | Does not capture SNV mutations or other copy-number-neutral events | [66] |