spliceJAC: transition genes and state-specific gene regulation from single-cell transcriptome data
- PMID: 36321549
- PMCID: PMC9627675
- DOI: 10.15252/msb.202211176
spliceJAC: transition genes and state-specific gene regulation from single-cell transcriptome data
Abstract
Extracting dynamical information from single-cell transcriptomics is a novel task with the promise to advance our understanding of cell state transition and interactions between genes. Yet, theory-oriented, bottom-up approaches that consider differences among cell states are largely lacking. Here, we present spliceJAC, a method to quantify the multivariate mRNA splicing from single-cell RNA sequencing (scRNA-seq). spliceJAC utilizes the unspliced and spliced mRNA count matrices to constructs cell state-specific gene-gene regulatory interactions and applies stability analysis to predict putative driver genes critical to the transitions between cell states. By applying spliceJAC to biological systems including pancreas endothelium development and epithelial-mesenchymal transition (EMT) in A549 lung cancer cells, we predict genes that serve specific signaling roles in different cell states, recover important differentially expressed genes in agreement with pre-existing analysis, and predict new transition genes that are either exclusive or shared between different cell state transitions.
Keywords: attractor linear stability; cell state transition; gene regulatory network; mRNA splicing; single-cell RNA sequencing.
© 2022 The Authors. Published under the terms of the CC BY 4.0 license.
Figures
![Figure 1](https://cdn.statically.io/img/www.ncbi.nlm.nih.gov/pmc/articles/instance/9627675/bin/MSB-18-e11176-g005.gif)
As input, spliceJAC requires unspliced and spliced mRNA count matrices as well as cell annotations.
spliceJAC fits the mRNA count data within each cell state to a multivariate mRNA splicing model.
The output of the model is a set of gene–gene interaction matrices that encode gene–gene interactions in each cell state.
The switch between two cell states is interpreted as a transition on a high‐dimensional landscape shaped by the underlying state‐specific gene regulatory networks.
By projecting the unstable eigenvalues of the inferred Jacobian matrix, spliceJAC predicts key transition genes (TG).
Each cell state exhibits specific interactions between genes that are captured in a regulatory network. Downstream analysis of the network identifies important signaling hubs such as strong regulators that modulate many other genes and strong targets that receive inputs from multiple genes.
A state transition coincides with a rearrangement of the state‐specific regulatory network.
![Figure 2](https://cdn.statically.io/img/www.ncbi.nlm.nih.gov/pmc/articles/instance/9627675/bin/MSB-18-e11176-g003.gif)
- A
The phase space of a bistable toggle switch including nullclines (silver lines), stable fixed points (blue dots), and stochastic perturbation around stable fixed points (red lines). X‐ and y‐coordinates represent unspliced mRNA counts of genes X and Y.
- B
Ground truth (top) and inferred (bottom) interaction matrices for the two stable fixed.
- C
The phase space of a monostable circuit of three genes that activate each other in a loop.
- D
Ground truth and inferred interaction matrix of the three genes circuit.
- E
The EMT circuit proposed by Tian et al (2013). Green and red nodes highlight epithelial and mesenchymal genes, while pointing and t‐shaped arrows represent activation or inhibition, respectively.
- F
Bifurcation diagram showing the available attractors as a function of TGF‐beta inducer. The red dotted line highlights a value leading to tristability used for spliceJAC testing thereafter.
- G, H
Ground truth (G) and estimated (H) interaction matrices in the mesenchymal state.
- I, J
Comparison with existing GRN inference methods based on the absolute difference between ground truth and prediction (I) and maximization of the fraction of correct matrix element signs (J). Green, orange, and red bars showcase results for the Epithelia, Hybrid E/M and mesenchymal states, respectively.
- K
The AUPRC ratio for all methods in the Beeline pipeline and spliceJAC. For each circuit state, the AUPRC scores are normalized by the average score achieved for the state. NA = no output file was generated. NA* = an output file without any predicted edge was generated.
![Figure EV1](https://cdn.statically.io/img/www.ncbi.nlm.nih.gov/pmc/articles/instance/9627675/bin/MSB-18-e11176-g009.gif)
- A
Left: The phase space of the toggle switch synthetic circuit. Right: The eigenvalues of the Jacobian matrices in the three fixed points, ranked in ascending order, inferred by spliceJAC.
- B–D
The ground truth (red) and inferred (blue) eigenvalue spectrum in the three fixed points of the EMT circuit.
- E
The inferred eigen‐spectrums cells sampled from multiple cell states are merged. Positive eigenvalues in the red circle indicate instability.
- F
Comparison of largest eigenvector inferred by spliceJAC in the three stable states and in the cell mixtures. The boxplot central bands and boxes depict average, first to third quantile (Q1–Q3) range, respectively, while the whisker extension corresponds to 1.5× the interquartile range (IQR). The boxplot results are computed over 10 independent simulations.
![Figure 3](https://cdn.statically.io/img/www.ncbi.nlm.nih.gov/pmc/articles/instance/9627675/bin/MSB-18-e11176-g002.gif)
UMAP embedding of the pancreas epithelium dataset from Bastidas‐Ponce et al (2019). Arrows indicate cell transitions identified with PAGA.
A core gene regulatory network governing gene expression in Ductal state. Node size depicts gene expression level within the Ductal cluster while the color scale depicts betweenness centrality of the node.
Scatterplot of the genes based on incoming interaction strength (x‐axis) and outgoing interaction strength (y‐axis).
State‐to‐state variability of gene betweenness centrality across cell states quantified by the interquartile range. Leftmost genes have highest state‐to‐state variability, whereas rightmost genes have lowest state‐to‐state variability.
A core GRN including the top Differentially Expressed Genes (DEG) of the Ngn3 high EP cell state and the top Transition Genes (TG) for the transition toward the Pre‐endocrine state.
Change in incoming and outgoing signaling scores during the transitioning from Ngn3 high EP to pre‐endocrine.
Flowchart highlighting the shared and unique Transition Genes (TG) for differentiation from Pre‐endocrine to Alpha, Beta, Delta, and Epsilon, respectively.
Gene Ontology analysis for the top five identified transitioning genes leading to the Alpha, Beta, Delta, and Epsilon cluster, respectively.
![Figure EV2](https://cdn.statically.io/img/www.ncbi.nlm.nih.gov/pmc/articles/instance/9627675/bin/MSB-18-e11176-g008.gif)
- A
The differential GRN between the Ngn3 high EP and Pre‐endocrine cell states. Node colormap indicates the gene expression fold‐change between the cell states.
- B
The top 10 differential interactions between the Ngn3 high EP and Pre‐endocrine cell states. Arrows depict the interaction strength change.
- C, D
The conserved GRN (C) and top 10 conserved interactions (D) highlight the interactions that maintained similar strength from the Ngn3 high EP to the Pre‐endocrine cell state.
![Figure EV3](https://cdn.statically.io/img/www.ncbi.nlm.nih.gov/pmc/articles/instance/9627675/bin/MSB-18-e11176-g004.gif)
The area under the precision‐recall curve (AUPRC) when predicting the GRN of the observed GRN (y‐axis) using the GRN of the predictor state (x‐axis).
The goodness of GRN prediction along a developmental trajectory. For each point, the red point shows the AUPRC between the GRNs of the starting and final state. For comparison, the blue dots and error bars show the average and standard deviation (SD) of AUPRC obtained when comparing the GRN of the starting state to the GRN of any other state in the dataset (n = 6 other states in the dataset excluding the final state).
![Figure 4](https://cdn.statically.io/img/www.ncbi.nlm.nih.gov/pmc/articles/instance/9627675/bin/MSB-18-e11176-g007.gif)
UMAP embedding of the A549 cells under TGFB induction(Cook & Vanderhyden, 2020). Arrows indicate cell transitions for partial and complete EMT.
Average expression of epithelial and mesenchymal genes in the three identified cell states.
Variability of total signaling across cell states quantified by the total signaling range across the three identified cell states. Leftmost genes have high state‐to‐state variability in their regulatory behavior, whereas rightmost genes have low state‐to‐state variability.
Detail for the top five variable genes identified in panel (C). The bar plot showcases the total signaling of the genes in each of the three cell states.
Same as (D) for the top five least variable genes.
Top 10 transitioning genes in the “partial EMT” transition from epithelial to hybrid E/M states.
Same as (F) for the “complete EMT” transition from hybrid E/M to mesenchymal states.
A core GRN indicating the inferred interactions between top differentially expressed genes (DEG) of the epithelial state and the top transition genes of partial EM Transition.
Same as (H) for the complete EMT transition.
Gene Ontology analysis for the top 10 identified transitioning genes for partial and complete EMT. The analysis compares transitioning genes that are exclusive to partial EMT, exclusive to complete EMT, and shared between the two transitions.
Similar articles
-
Spatial transition tensor of single cells.Nat Methods. 2024 Jun;21(6):1053-1062. doi: 10.1038/s41592-024-02266-x. Epub 2024 May 16. Nat Methods. 2024. PMID: 38755322 Free PMC article.
-
Comparative gene co-expression network analysis of epithelial to mesenchymal transition reveals lung cancer progression stages.BMC Cancer. 2017 Dec 6;17(1):830. doi: 10.1186/s12885-017-3832-1. BMC Cancer. 2017. PMID: 29212455 Free PMC article.
-
Roles and Regulation of Epithelial Splicing Regulatory Proteins 1 and 2 in Epithelial-Mesenchymal Transition.Int Rev Cell Mol Biol. 2016;327:163-194. doi: 10.1016/bs.ircmb.2016.06.003. Epub 2016 Jul 30. Int Rev Cell Mol Biol. 2016. PMID: 27692175 Review.
-
High-throughput mRNA and miRNA profiling of epithelial-mesenchymal transition in MDCK cells.BMC Genomics. 2015 Nov 16;16:944. doi: 10.1186/s12864-015-2036-9. BMC Genomics. 2015. PMID: 26572553 Free PMC article.
-
Complex changes in alternative pre-mRNA splicing play a central role in the epithelial-to-mesenchymal transition (EMT).Semin Cancer Biol. 2012 Oct;22(5-6):417-27. doi: 10.1016/j.semcancer.2012.04.003. Epub 2012 Apr 23. Semin Cancer Biol. 2012. PMID: 22548723 Free PMC article. Review.
Cited by
-
Spatial transition tensor of single cells.Nat Methods. 2024 Jun;21(6):1053-1062. doi: 10.1038/s41592-024-02266-x. Epub 2024 May 16. Nat Methods. 2024. PMID: 38755322 Free PMC article.
-
Data- and theory-driven approaches for understanding paths of epithelial-mesenchymal transition.Genesis. 2024 Apr;62(2):e23591. doi: 10.1002/dvg.23591. Genesis. 2024. PMID: 38553870 Review.
-
Data-driven energy landscape reveals critical genes in cancer progression.NPJ Syst Biol Appl. 2024 Mar 8;10(1):27. doi: 10.1038/s41540-024-00354-4. NPJ Syst Biol Appl. 2024. PMID: 38459043 Free PMC article.
-
Order-of-Mutation Effects on Cancer Progression: Models for Myeloproliferative Neoplasm.Bull Math Biol. 2024 Feb 16;86(3):32. doi: 10.1007/s11538-024-01257-5. Bull Math Biol. 2024. PMID: 38363386 Free PMC article.
-
Quantifying Landscape-Flux via Single-Cell Transcriptomics Uncovers the Underlying Mechanism of Cell Cycle.Adv Sci (Weinh). 2024 Apr;11(16):e2308879. doi: 10.1002/advs.202308879. Epub 2024 Feb 14. Adv Sci (Weinh). 2024. PMID: 38353329 Free PMC article.
References
-
- Akers K, Murali TM (2021) Gene regulatory network inference in single‐cell biology. Curr Opin Syst Biol 26: 87–97
-
- Aubin‐Frankowski P‐C, Vert J‐P (2020) Gene regulation inference from single‐cell RNA‐seq data with linear differential equations and velocity inference. Bioinformatics 36: 4774–4780 - PubMed
-
- Bastidas‐Ponce A, Tritschler S, Dony L, Scheibner K, Tarquis‐Medina M, Salinno C, Schirge S, Burtscher I, Böttcher A, Theis FJ et al (2019) Comprehensive single cell mRNA profiling reveals a detailed roadmap for pancreatic endocrinogenesis. Development 146: dev173849 - PubMed
Publication types
MeSH terms
Substances
Grants and funding
LinkOut - more resources
Full Text Sources