Advanced Analysis
Overview
This section summarizes common advanced single-cell transcriptome analysis methods, aiming to help researchers extract richer biologically interpretable conclusions. We focus on practicality, highlighting the definition, applicable scenarios, platform parameters, and key points for result interpretation for each method, and providing application examples, common pitfalls, and optimization suggestions to facilitate data analysis on the cloud platform.
TIP
How to Read This Section
- Since this section is long, we recommend using the table of contents on the right to quickly jump to topics of interest and expand specific content as needed.
Immune Repertoire Analysis
Definition
Immune Repertoire (IR) refers to the collection of all T-cell receptors (TCRs) and B-cell receptors (BCRs) in an individual's adaptive immune system at a specific time point. Its diversity stems from V(D)J gene rearrangement during lymphocyte development. As an important component of single-cell research, immune repertoire analysis comprehensively assesses the status and dynamics of the immune system through high-throughput sequencing of TCRs and BCRs. By resolving key CDR3 sequences, this analysis can identify and quantify massive T/B cell clones, thereby revealing the characteristics of adaptive immune responses in health or disease states (such as infection, tumors, autoimmune diseases, etc.).
View More (Expand/Collapse)
Significance
- Reveal Immune Diversity: Assess the health status of the immune system and its potential ability to respond to antigens. Reduced diversity may indicate impaired immune function or clonal selection for specific antigens.
- Track Clonal Dynamics: By comparing samples from different time points, tissues, or conditions, specific T/B cell clone expansion, contraction, or persistence can be monitored to understand the dynamic process of immune response.
- Identify Disease-Associated Clones: In tumor immunotherapy, infectious diseases, and autoimmune disease research, identifying specific TCR/BCR clones associated with disease progression or treatment response provides clues for developing diagnostic markers and targeted therapies.
- Understand Immune Response Mechanisms: Combining single-cell transcriptome data, clonotype information can be linked to the gene expression profile of host cells to deeply explore the functional status of different clones (such as effector, memory, exhaustion, etc.).
- Evaluate Vaccine and Immunotherapy Effects: By analyzing changes in the immune repertoire before and after vaccination or treatment, the effectiveness of immune interventions can be assessed, and their mechanisms of action revealed.
Workflow and Interpretation
Pseudotime Analysis
Definition
Pseudotime analysis is a computational method used to infer the relative progression of cells along a dynamic biological process (e.g., development, differentiation, disease progression, or response to stimuli) from single-cell transcriptomic data. By measuring transcriptomic similarity between cells, it maps discrete cell samples into a low-dimensional space and orders them according to their intrinsic continuity, assigning each cell a relative scale (pseudotime value) representing "progression extent." It is important to emphasize that pseudotime is not equivalent to real time but reflects the relative position of cell states along an assumed trajectory.
Typical analysis workflows include: selecting ordering genes that represent the process, learning or fitting a principal graph structure (such as a tree or graph model) in low-dimensional space, and projecting cells onto this principal graph to calculate pseudotime values and identify branch points and terminal states.
View More (Expand/Collapse)
Significance
- Overcome Cell Asynchrony: Single-cell samples are usually composed of mixtures of cells at different stages of development or response. Pseudotime analysis can reorder these asynchronous cells according to their intrinsic process, revealing continuous biological changes.
- Reconstruct Continuous Processes: Compared to discrete clustering, pseudotime can display continuous transition trajectories between cell states, helping to understand gradual changes from initial to terminal states.
- Identify Transitional States and Rare Cells: It can detect intermediate state cells or rare transition cells between two stable states, which often carry key fate determination information.
- Discover Dynamic Regulatory Genes and Modules: By analyzing gene expression changes along pseudotime, key regulatory factors upregulated or downregulated at different stages of the process and co-varying gene modules can be identified, providing candidate gene sets for subsequent mechanistic studies.
- Reveal Branch Points and Fate Decisions: When the trajectory contains branch structures, pseudotime analysis can locate fate divergence points and compare gene expression differences between branches, thereby proposing verifiable hypotheses about cell fate choices.
- Guide Experimental Design and Multi-Method Validation: Pseudotime results can guide validation experiments such as time-series sampling, intervention experiments, or lineage tracing; they can also be combined with other methods (such as CytoTRACE based on transcriptional diversity or scVelo based on RNA velocity) to enhance conclusion reliability.
TIP
Pseudotime analysis is particularly suitable for studying biological problems with continuous change processes, such as:
- Stem cell differentiation and development.
- Immune cell activation and exhaustion.
- Disease (e.g., cancer) occurrence and evolution.
- Cell response processes to drugs or environmental stimuli.
Methods and Selection Guide
| Tool | Core Principle | Pros | Cons | Recommended Scenario |
|---|---|---|---|---|
| Monocle 2 | Pseudotime + DDRTree Algorithm. Constructs a minimum spanning tree of cells in low-dimensional space through dimensionality reduction and graph learning; a classic method for parsing complex branches. | - Stable and Classic: Constructed branch trajectories are clear and validated by extensive literature. - Mature Ecology: Supported by numerous tutorials and papers. | - Computational Performance: Slow for very large datasets (>100k cells), high memory consumption. - Start Point Dependency: Requires user to specify trajectory start point. | Top Recommendation. When you want to clearly demonstrate how cells differentiate from one state to multiple different terminal states, Monocle 2 is a proven, reliable choice. |
| Monocle 3 | Pseudotime + UMAP Embedding. Learns graph structure directly on dimensionality reduction plots like UMAP to infer trajectories; a more simplified workflow and newer strategy. | - Fast and Scalable: Friendly to massive datasets, capable of handling millions of cells. - Good Compatibility: Good compatibility with mainstream ecosystems like Scanpy. | - Still Developing: Algorithm is iterating rapidly; parsing ability for complex branches is sometimes less stable and clear than Monocle 2. | Efficient alternative when handling massive datasets or when trajectory structure is relatively simple (e.g., linear or single branch). |
| CytoTRACE | Transcriptional Diversity. Based on the core hypothesis: cells with higher differentiation potential express more genes; cells with higher differentiation degrees have more specialized gene expression patterns. | - Completely Unbiased: No need to specify a start point; automatically predicts cell differentiation potential. - Good at Root Finding: Very powerful in identifying the "root" (most primitive cell group) of the trajectory. | - No Trajectory Graph: It mainly provides a high-to-low differentiation potential ranking (a value) rather than a visual path graph. | Strongly recommended when you are unsure which cell group is the start point or want to objectively verify if the start point selection for tools like Monocle is correct. |
| scVelo | RNA Velocity. Infers the transcriptional dynamic direction of each cell in the "future" few hours by quantifying the abundance of pre-mRNA (unspliced) and mature-mRNA (spliced). | - Predict "Future": Reveals instantaneous direction and rate of cell state transitions, providing real dynamic information. - Reveal Cyclic Processes: Very effective for depicting cyclic processes like the cell cycle. | - High Data Requirements: Requires high-quality data capable of effectively capturing intron reads. - Complex Interpretation: Velocity vector fields can be messy and require careful interpretation. | Best choice when you want to know the "direction" and "velocity" of cell state transitions, not just the "path". Also suitable for exploring dynamic equilibrium systems. |
Summary:
- Standard Workflow: Monocle 2 (depict path) + CytoTRACE (determine start point).
- Explore Direction and Rate: If data quality permits, use scVelo for deeper dynamic analysis.
- Massive Datasets: Prioritize Monocle 3.
Cell-Cell Communication Analysis
Definition
Cell-cell communication analysis infers signal exchange between cell populations via ligand-receptor (L-R) pairs and downstream signaling pathways from single-cell/spatial transcriptomic data. Its goal is not only to identify possible L-R pairs but also to quantify "sender/receiver" roles and pathway strengths at the pathway and system levels, providing candidate signaling molecules and mechanistic hypotheses for subsequent experimental validation.
TIP
Cell communication results are "predictive" rather than "causal" conclusions. Be sure to combine them with expression evidence, literature, and experimental validation (flow cytometry, in situ, or functional blocking experiments).
View More (Expand/Collapse)
Common Software and Differences
| Tool | Core Principle | Pros | Cons | Recommended Scenario |
|---|---|---|---|---|
| CellChat (R) | Based on CellChatDB, aggregates multisubunit and cofactor info, combines network analysis and pathway summary to quantify communication strength and cell roles. | Rich visualization (circle/chord/sankey, etc.), pathway summary, and sender/receiver role analysis; suitable for system-level comparison. | Relies on prior database coverage; high computational resource requirement for massive datasets; sensitive to low expression/rare cells. | First choice for system-level pathway comparison, sender/receiver role identification, and pathway prioritization. |
| CellPhoneDB (Python/CLI) | Uses subunit minimum expression principle to represent complex expression, employs cell label permutation test for significance. | Sensitive to complexes and rigorous statistical method; mature database for human data. | Official focus is HUMAN (other species need mapping); relatively limited visualization and pathway summary. | First choice for strict LR significance screening and conservative result validation in human data. |
| NicheNet (R) | Integrates ligand-receptor, signaling, and TF→target gene networks, uses network propagation to assess ligand regulatory potential on target genes. | Can provide mechanistic prediction of ligand→target gene and ligand activity scoring, suitable for explaining differential expression in receptor cells. | Does not directly provide sender/receiver strength metrics between groups; usually requires differential genes as input. | First choice when the focus is on whether a ligand can explain differential genes and downstream regulatory mechanisms in receptor cells. |
Note: Differences in database coverage, complex annotation, and statistical strategies between tools can significantly affect detection results; cross-tool validation is recommended when reporting candidates.
Recommendations
- For system-level visualization and pathway comparison (including sender/receiver analysis): Prioritize
CellChat(excellent visualization and pathway summary capabilities). - For human data with a focus on statistical significance of L-R pairs:
CellPhoneDBis a robust choice (emphasizes subunit minimum expression strategy and permutation tests). - For explaining differential genes in receptor cells via ligands: Use
NicheNetto generate candidate lists of ligand→target genes and activity scores. - Comprehensive Strategy (Recommended): After software analysis, validate using wet lab experiments or spatial data.
Common Pitfalls
WARNING
- Do not rely solely on "strength values" from a single software to draw experimental conclusions; different methods measure different meanings.
- Avoid directly comparing values between different groups without unified annotation and preprocessing; such comparisons are prone to technical bias.
References
- CellChat Official Repository:
https://github.com/sqjin/CellChat - CellPhoneDB Official Repository:
https://github.com/ventolab/CellphoneDB - NicheNet Official Repository:
https://github.com/saeyslab/nichenetr
Copy Number Variation (CNV) Analysis
Definition
Copy Number Variation (CNV) is a type of genomic structural variation referring to the increase or decrease in copy number of DNA segments larger than 1kb compared to a reference genome. CNV is a major source of genomic difference within and between species and is closely related to the development of many diseases, especially in tumor research, where CNV is an important feature of tumor heterogeneity.
In single-cell RNA sequencing (scRNA-seq) data, CNV can be indirectly inferred by analyzing gene expression changes across continuous genomic regions. If the overall gene expression level in a chromosomal region increases or decreases, it may imply an increase or decrease in copy number in that region.
View More (Expand/Collapse)
Significance
Single-cell CNV analysis is extremely important in tumor research:
- Distinguish Tumor and Non-Tumor Cells: Tumor cells are usually accompanied by extensive genomic instability, manifesting as widespread CNVs. Analyzing CNV patterns of single cells can effectively distinguish malignant tumor cells from normal cells (such as immune cells, stromal cells) mixed in tumor tissues.
- Reveal Tumor Clonal Heterogeneity: Tumor tissues consist of sub-clones with different genomic characteristics. Single-cell CNV analysis can identify tumor sub-clones with different CNV patterns, revealing the clonal structure and evolutionary relationship within the tumor.
- Identify Key Driver Genes: CNV analysis can locate chromosomal regions with frequent copy number gains or losses, which may contain key oncogenes or tumor suppressor genes, providing clues for finding new therapeutic targets.
- Assess Tumor Evolution and Drug Resistance: By comparing tumor cell CNV patterns at different treatment stages or metastatic sites, tumor evolutionary paths can be tracked, and genomic variations related to drug resistance can be studied.
InferCNV vs. CopyKAT
InferCNV and CopyKAT are two widely used bioinformatics tools for inferring CNV from scRNA-seq data. They differ in algorithm principle, applicable scenarios, and result focus.
| Tool | Core Principle | Pros | Cons | Recommended Scenario |
|---|---|---|---|---|
| InferCNV | Infers CNV by comparing gene expression profiles of tumor cells with a set of "normal" reference cells using a sliding window to smooth gene expression. | - Classic algorithm, reliable results - Focuses on large-scale chromosomal arm-level variations | - Must provide high-quality normal reference cells - Relatively lower resolution - Computationally intensive, time-consuming - Results require user interpretation and cell grouping | Suitable for scenarios with clear, reliable normal cell references (e.g., adjacent normal tissue, immune cells) and a focus on large-scale variations. |
| CopyKAT | Uses an integrated Bayesian method to identify CNVs by comparing with a mixture model of genomic position and expression. | - No need for predefined reference cells, automatic identification - Higher resolution (~5Mb) - High computational efficiency, fast - Automatically predicts malignant/normal status of cells | - Accuracy relies on sufficient normal cells in data as internal reference - Automatic sub-clone division might be too simplified for complex intra-tumor heterogeneity | Suitable for scenarios without clear normal controls or when automatic differentiation of tumor/normal cells is desired. Its automation makes it preferred for exploratory analysis. |
Selection Advice:
- If your data contains clear, reliable normal cells as reference (e.g., non-epithelial cells from adjacent tissues, or clearly annotated immune cells), and you are more interested in large-scale chromosomal arm-level variations, InferCNV is a classic and reliable choice.
- If your data lacks clear normal cell controls, or you want the algorithm to automatically distinguish tumor and normal cells, and you are interested in higher resolution CNV events, CopyKAT is a better choice. Its higher degree of automation makes results easier to interpret.
- In actual analysis, both tools can be used simultaneously to cross-validate results for more reliable conclusions.
Regulatory Network Analysis
Definition
Regulatory network analysis infers gene regulatory relationships and identifies functional modules from single-cell transcriptomic data. By integrating gene expression correlation, transcription factor binding site information, and cell-type-specific expression patterns, it constructs cell-specific gene regulatory networks, revealing transcriptional regulatory mechanisms driving cell state transitions and functional differentiation.
Typical analysis workflows include: identifying co-regulated gene modules based on gene co-expression patterns (e.g., hdWGCNA), inferring regulatory relationships between transcription factors and target genes (e.g., SCENIC), and revealing the biological significance of modules through functional enrichment analysis.
View More (Expand/Collapse)
Significance
- Reveal Transcriptional Regulatory Mechanisms: Identifying key transcription factors and their target genes reveals mechanisms behind cell state transitions.
- Identify Functional Gene Modules: Gene co-expression network analysis identifies gene modules co-expressed in specific cell types, providing clues for understanding cell functions.
- Discover Cell-Type-Specific Regulators: Analyzing regulator activity differences across cell types identifies key cell-type-specific regulators.
- Understand Molecular Basis of Heterogeneity: Helps understand gene regulatory differences behind cell heterogeneity, providing molecular basis for cell classification and annotation.
- Guide Target Screening: Identifying key regulatory modules and factors provides candidate gene sets for subsequent functional validation and drug target screening.
- Integrate Multi-Omics: Can integrate gene expression, epigenetics, and protein interaction data for comprehensive regulatory mechanism parsing.
TIP
Regulatory network analysis is particularly suitable for studying biological problems with complex regulatory relationships, such as:
- Molecular mechanisms of cell fate determination.
- Regulatory network rewiring of disease-associated genes.
- Drug mechanism and resistance research.
- Dynamic changes in transcriptional regulation during development.
Methods and Selection Guide
| Tool | Core Principle | Pros | Cons | Recommended Scenario |
|---|---|---|---|---|
| hdWGCNA | Gene Co-expression Network + Module Identification. Constructs weighted gene co-expression networks based on correlation, identifies functional modules via hierarchical clustering. | - Modular Analysis: Identifies cell-type-specific functional modules - Eigengene Extraction: Quantifies module activity via Module Eigengenes - Functional Enrichment: Reveals module biology via GO/KEGG | - Indirect Regulation: Based on co-expression, cannot directly verify regulation - Parameter Sensitive: Results affected by soft thresholding parameters | Top Recommendation. Mature and reliable choice for identifying co-expressed gene modules in cell types and understanding their functions. |
| SCENIC | Network Inference + Regulon Activity. Infers TF-target relationships from co-expression, verifies with motif analysis, calculates regulon activity via AUCell. | - Direct Regulation: Infers direct TF-target relationships - Activity Quantification: Quantifies regulon activity in each cell via AUC - State Identification: Groups/annotates cells based on network activity | - Computationally Complex: Complex workflow, high resource demand - High Data Quality: Requires high input data quality | Best choice for deeply understanding transcriptional regulatory mechanisms and identifying key TFs and their targets. |
Gene Set Scoring Analysis
Definition
Gene set scoring evaluates the activity of predefined gene sets in single-cell transcriptomic data. By comprehensively scoring the activity for each cell or cell group, this method quantifies the enrichment of specific biological pathways, functions, or states. This scoring mechanism provides a powerful tool for revealing intrinsic cell heterogeneity and deeply exploring biological differences across cell states.
View More (Expand/Collapse)
Significance
- Functional Annotation: Annotate unknown cell groups to reveal their biological roles.
- State Comparison: Compare activity changes of specific pathways under different conditions or cell types.
- Heterogeneity Exploration: Discover functional heterogeneity within the same cell group.
- Biological Insight: Provide molecular explanations for disease mechanisms, differentiation, drug response, etc.
Methods and Selection Guide
| Tool | Core Principle | Pros | Cons | Recommended Scenario |
|---|---|---|---|---|
| scMetabolism | Based on VISION and AUCell, built-in 78 KEGG/REACTOME metabolic pathways. | Optimized for metabolism, intuitive results. | Limited to 78 pathways, not for other sets. | Metabolic Pathway Focus. Recommended for quick study of metabolic reprogramming. |
| GSVA | Non-parametric, unsupervised method converting expression matrix to enrichment score matrix. | Broad applicability, supports custom/public sets; good for cross-sample comparison. | Interpretation depends on set quality; may be less sensitive to sparse data. | General Gene Set Variation. Recommended for comprehensive, unbiased analysis using large DBs like MSigDB. |
| Scoring | Integrates AUCell, UCell, singscore, AddModuleScore. | High flexibility, supports custom sets and cross-validation. | Requires user understanding of algorithms. | Flexible Tool. Recommended for custom sets or cross-validating with multiple algorithms. |
| Module Score | AddModuleScore | Uses Seurat's AddModuleScore function. | - Simple, intuitive, adjustable plotting - May be affected by set size/expression | Suitable for quick assessment and visualization of gene set activity. |
Perturbation Analysis
Definition
Perturbation analysis in single-cell transcriptomics assesses transcriptional response differences and sensitivity of cell populations or types under defined external/internal conditions (e.g., disease, drug, knockout, timepoint). The goal is to identify cell types or gene modules most responsive to conditions, revealing potential biological regulation and functional differences.
View More (Expand/Collapse)
Significance
- Identify Key Responsive Cell Types: Find cell types most sensitive to perturbation for functional validation.
- Reveal Condition-Specific Mechanisms: Identify genes/pathways changed under perturbation to understand regulation.
- Prioritize Resources: Prioritize cell subsets for in-depth validation in complex tissues.
- Guide Intervention/Biomarkers: Provide basis for precision therapy and biomarker development.
Setup
- Perturbation Factor: Corresponds to a
metadatacolumn (e.g.,treatment,disease_status). Ensure clear, mutually exclusive labels (e.g., treated vs control). - Perturbation Objects: Target groups to compare (whole sample, specific cell type). Ensure sufficient cells (>100/group) and consider batch effects.
TIP
Workflow: Confirm factor/groups, then run analysis on biologically relevant cell types to get robust prioritization.
- Tool: Use Augur for prioritization; combine with differential expression, regulatory network, cell communication for cross-validation.
GeneActivity Analysis
Definition of GeneActivity Analysis
GeneActivity analysis infers potential transcriptional activity by calculating chromatin accessibility in gene bodies and their upstream regulatory regions, to fill the information gap of "only open regions, no gene expression" in single-cell ATAC data. In single-cell multi-omics research, scATAC-seq data provides information on chromatin openness but cannot directly reflect gene transcriptional activity. GeneActivity analysis obtains activity values through counting the number of fragments in the promoter (usually 2 kb upstream of TSS) and gene body regions of each gene, followed by normalization, thereby inferring whether the gene has transcriptional potential.
The difference between GeneActivity and Gene Expression is:
- GeneActivity: Based on scATAC-seq data, reflects chromatin openness and potential transcriptional capacity
- Gene Expression: Based on scRNA-seq data, reflects actual mRNA abundance
The two are usually positively correlated, but inconsistencies may arise due to epigenetic silencing, post-transcriptional regulation, or technical noise. Comparing the two can provide deeper understanding of regulatory mechanisms.
View More Content (Expand/Collapse)
Significance of GeneActivity Analysis
GeneActivity analysis has important biological significance in single-cell multi-omics research:
- Inferring potential transcriptional activity: By counting chromatin accessibility signals in gene bodies and their upstream regulatory regions, the potential transcriptional activity of each gene can be inferred, providing a foundation for gene functional interpretation. This is particularly useful when only ATAC data is available without RNA data.
- Assisting cell type annotation: GeneActivity information can be used to assist cell type identification and annotation, filling the gap of gene expression information lacking in ATAC data, making cell type annotation based on ATAC data more accurate.
- Identifying differentially accessible genes: By comparing GeneActivity values across different cell populations (such as cell types, states, treatment groups, etc.), regulatory genes with significantly high or low activity in specific populations can be identified, similar to differential expression analysis.
- Integrating multi-omics information: Combining scRNA-seq and scATAC-seq data allows comparison of heterogeneity and correlation between RNA expression and ATAC gene accessibility at different cell population levels. Spearman rank correlation coefficients can be used to assess similarities between cell populations, revealing consistency or decoupling relationships between epigenetic regulation and transcriptional expression.
- Revealing transcriptional regulatory mechanisms: By analyzing the consistency and differences between gene activity and gene expression, special regulatory phenomena such as "only open but not expressed" or "only expressed but not open" can be discovered, revealing the complexity of epigenetic regulation.
TIP
GeneActivity analysis is particularly suitable for studying the following biological questions:
- Cell type annotation based on ATAC data
- Identifying GeneActivity differences between different cell populations
- Integrating multi-omics data to reveal relationships between epigenetic regulation and transcriptional expression
- Discovering potential key regulatory genes and regulatory mechanisms
Tool recommendation: Use GeneActivity to complete GeneActivity calculation, differential analysis, and multi-omics correlation analysis.
ATAC Differential Accessibility Analysis
Definition of ATAC Differential Accessibility and Regulatory Network Analysis
ATAC Differential Accessibility and Regulatory Network Analysis (ATAC_DiffEnrich) is a differential accessibility and regulatory network analysis method specifically designed for single-cell multi-omics (ATAC + RNA) data. Its core goal is to identify differentially open chromatin regions between cell populations or experimental groups, and trace back to core transcription factors (TF) through "multiple evidence chain closure" strategy, while mapping target genes and functional pathways downstream.
In single-cell multi-omics research, changes in chromatin accessibility directly determine whether TF can bind to cis-regulatory elements (enhancers, promoters, etc.), thereby initiating or inhibiting downstream gene transcription. Compared to observing only RNA expression changes, ATAC-level signals are closer to the "starting point" of regulation.
ATAC_DiffEnrich analysis not only helps locate differentially open regions but also systematically analyzes regulatory networks through the following strategies:
- Tracing upstream: Starting from differentially open regions, completing the evidence chain of "enrichment-activity-expression-physical binding" through Motif enrichment → chromVAR activity → expression correlation → Footprint physical confirmation.
- Validating downstream: Mapping differentially open regions to nearest target genes and performing GO/KEGG functional enrichment to translate epigenetic changes into specific biological processes and signaling pathways.
View More Content (Expand/Collapse)
Significance of ATAC Differential Accessibility and Regulatory Network Analysis
ATAC_DiffEnrich analysis has important biological significance in single-cell multi-omics research:
- Differential accessibility region identification: Through Wilcoxon rank-sum test, identify significantly different open chromatin regions (DiffPeaks) between cell populations or experimental groups, filtering threshold conditions that meet significance and effect size.
- Motif enrichment and activity analysis: Perform Motif enrichment analysis on differential Peaks to identify potential upstream TF; use chromVAR algorithm to calculate Motif activity (Deviation Score) at single-cell level and perform differential TF activity analysis.
- Multi-dimensional confirmation of core TF: Complete multi-dimensional evidence chain from sequence enrichment to physical binding through global correlation analysis of TF activity and RNA expression, Venn diagram intersection screening, and Tn5 footprint analysis (Footprint).
- Target gene functional mapping: Map differential Peaks to nearest genes and perform GO/KEGG enrichment analysis to translate chromatin accessibility changes into interpretable biological conclusions.
TIP
ATAC_DiffEnrich analysis is particularly suitable for studying the following biological questions:
- Identifying cell type-specific or condition-specific differential accessibility regions
- Analyzing causal relationships between epigenetic regulation and transcriptional expression
- Discovering key transcription factors driving cell state transitions
- Constructing complete TF→Peak→Gene regulatory chains
Tool recommendation: Use ATAC_DiffEnrich to complete ATAC differential accessibility and regulatory network analysis.
Peak2Gene Analysis
Definition of Peak2Gene Analysis
Peak2Gene (Peak-Gene linking) analysis is a method specifically designed for single-cell multi-omics (ATAC + RNA) data. Its core goal is to identify significant regulatory relationships between gene expression and nearby chromatin accessibility peaks. Peak2Gene analysis calculates the correlation between gene expression and ATAC signal intensity of nearby peaks in each cell, and uses generalized linear models to correct technical biases such as GC content, peak length, and distance, thereby inferring which peaks may participate in regulating which genes.
The difference between Peak2Gene analysis and traditional distance annotation:
- Peak2Gene analysis: Based on single-cell level correlation analysis, uses generalized linear models to correct technical biases, can identify cell type-specific peak-gene associations, providing more reliable regulatory evidence
- Traditional distance annotation: Based only on linear distance annotation, cannot consider actual correlation evidence and technical biases, with lower accuracy
Peak2Gene analysis not only helps reveal cis-regulatory networks (such as enhancer-gene connections) but can also be combined with motif enrichment/activity results to systematically mine direct target genes and regulatory effects of key transcription factors, making it one of the fundamental tools for multi-omics regulatory mechanism research.
View More Content (Expand/Collapse)
Significance of Peak2Gene Analysis
Peak2Gene analysis has important biological significance in single-cell multi-omics research:
- Identifying peak-gene regulatory relationships: By calculating the correlation between gene expression and accessibility signals of nearby peaks, identify peak sets that may regulate the gene, overcoming the limitations of relying solely on linear distance annotation and providing more accurate regulatory relationship inference.
- Constructing cis-regulatory networks: Helps reveal cis-regulatory networks (such as enhancer-gene connections), providing a foundation for multi-omics regulatory mechanism research, and is a key step in analyzing gene regulatory networks (GRN).
- Correcting technical biases: Uses generalized linear models to correct multiple technical biases such as GC content, overall accessibility, and peak length to obtain more reliable peak-gene associations, improving the accuracy of analysis results.
- Integrating multi-omics information: Combining motif enrichment/activity results, systematically mine direct target genes and regulatory effects of key transcription factors, construct complete TF→peak→gene regulatory axes, and achieve three-dimensional integration of "TF expression × motif-peak binding × target gene expression."
- Revealing cell type-specific regulation: Can identify cell type-specific peak-gene association patterns at each cell resolution, reveal regulatory differences between different cell types, and deeply understand cell fate determination mechanisms.
TIP
Peak2Gene analysis is particularly suitable for studying the following biological questions:
- Constructing gene regulatory networks (GRN), identifying TF→peak→gene regulatory axes
- Identifying cis-regulatory elements such as enhancer-gene connections
- Revealing cell type-specific regulatory mechanisms
- Integrating motif analysis to systematically mine direct target genes of key transcription factors
- Understanding causal relationships between epigenetic regulation and transcriptional expression
Tool recommendation: Use Peak2Gene to complete peak-gene association identification, correlation analysis, and regulatory network construction.
TF–Peak–Gene Trios Analysis
Definition of TF–Peak–Gene Trios Analysis
TF–Peak–Gene Trios analysis is the core functional unit in single-cell multi-omics regulatory network research. Its goal is: under specific cell states, systematically identify "which TF" through binding to "which regulatory Peak" ultimately regulates "which Gene", thereby reconstructing biologically interpretable TF → Peak → Gene regulatory logical chains.
TF–Peak–Gene Trios integrates molecular evidence from three levels into a complete regulatory unit:
- TF (regulatory initiator): Transcription factors initiate specific regulation by recognizing and binding to characteristic Motif motifs on target DNA.
- Peak (regulatory mediator): Represents open chromatin regions, often enriched in cis-regulatory elements such as enhancers and promoters; its accessibility is the physical prerequisite for TF function.
- Gene (functional output): Gene expression changes are the ultimate manifestation of upstream regulatory events at the transcriptional level.
View More Content (Expand/Collapse)
Significance of TF–Peak–Gene Trios Analysis
TF–Peak–Gene Trios analysis has important biological significance in single-cell multi-omics research:
- Integrating multi-dimensional evidence to construct regulatory chains: Integrates multi-dimensional evidence from four levels: Peak2Gene co-variation links, Motif–Peak physical binding evidence, TF expression levels, and TF chromatin activity (chromVAR), constructing high-confidence TF → Peak → Gene regulatory logical chains through strict multi-dimensional threshold filtering.
- Identifying cell type-specific regulatory relationships: Through integrating multi-omics data, can identify key TF-Peak-Gene Trios active in specific cell types or states, revealing regulatory logic of cell fate determination.
- Reducing false positives: Compared to methods based only on co-expression or only on Peak2Gene correlation, Trios analysis introduces multiple independent validations, significantly reducing false positive connections.
- Guiding experimental validation: By constructing complete regulatory chains, provides clear candidate targets for subsequent functional validation experiments.
TIP
TF–Peak–Gene Trios analysis is particularly suitable for studying the following biological questions:
- Reconstructing cell type-specific gene regulatory networks
- Identifying key transcription factors and their target genes driving cell state transitions
- Understanding causal relationships between epigenetic regulation and transcriptional expression
- Providing candidate regulatory relationships for functional validation experiments
Tool recommendation: Use Trios to complete TF-Peak-Gene Trios identification and validation.
SCENIC+ Regulatory Network Analysis
Definition of SCENIC+ Regulatory Network Analysis
SCENIC+ (Single-Cell Regulatory Network Inference and Clustering Plus) is a transcriptional regulatory network inference framework integrating scRNA-seq and scATAC-seq data. Its core advantage lies in using chromatin accessibility information to constrain TF–target gene relationship inference, significantly reducing false positive connections compared to traditional SCENIC methods, and constructing complete "transcription factor → cis-regulatory element → target gene" regulatory networks.
In single-cell multi-omics research, inferring TF–target gene relationships relying only on gene co-expression often results in a large number of false positives. SCENIC+ integrates chromatin accessibility information from scATAC-seq, validating TF regulatory relationships from two independent omics levels (RNA expression + ATAC accessibility), greatly improving the reliability of regulatory networks.
View More Content (Expand/Collapse)
Significance of SCENIC+ Regulatory Network Analysis
SCENIC+ regulatory network analysis has important biological significance in single-cell multi-omics research:
- Regulatory network inference: Combines dual evidence from gene co-expression (RNA) and chromatin accessibility (ATAC) to infer regulatory relationships between transcription factors and target genes, constructing high-confidence eRegulon (extended Regulon).
- eRegulon quality assessment and filtering: Through multi-dimensional indicators such as TF expression–AUC correlation and Gene-based and Region-based AUC consistency, automatically filters high-quality regulatory relationships and eliminates false positive connections.
- Regulon Specificity Score (RSS): Calculates the specificity enrichment degree of each eRegulon in various cell types, identifies cell type-specific regulatory factors, and helps analyze the regulatory logic of cell fate determination.
- Multi-modal joint visualization: Performs UMAP/t-SNE dimensionality reduction based on eRegulon AUC matrices, maps TF expression, Gene AUC, and Region AUC to the same base map, and provides combined heatmap-dot plot displays to comprehensively present the structure and dynamics of regulatory networks.
TIP
SCENIC+ regulatory network analysis is particularly suitable for studying the following biological questions:
- Constructing high-confidence gene regulatory networks (GRN)
- Identifying cell type-specific key transcription factors
- Analyzing regulatory mechanisms of cell state transitions
- Discovering disease-related regulatory network abnormalities
Tool recommendation: Use SCENICplus to complete SCENIC+ regulatory network analysis.
ATAC Pseudotime Analysis
Definition of ATAC Pseudotime Analysis
The continuity of chromatin accessibility patterns in scATAC-seq data is used to identify continuous paths of cells in epigenetic space through dimensionality reduction and graph learning, constructing directed developmental trajectories and calculating pseudotime values based on specified starting cells. Single-cell multi-omics data contains two dimensions of information: scRNA-seq and scATAC-seq. The temporal dynamic changes in chromatin accessibility often precede transcription, providing earlier indications of key regulatory events than expression, making it suitable for mining regulatory mechanisms and fate determination-related nodes—emphasizing the "cause" perspective.
View More Content (Expand/Collapse)
Significance of ATAC Pseudotime Analysis
Cell differentiation is a continuous process involving gradual changes in epigenetic states. At single-cell resolution, we can observe:
- Revealing the continuity of cell differentiation: Cells at different differentiation stages coexist in samples. Pseudotime analysis can reorder these asynchronous cells according to their intrinsic differentiation processes, revealing continuous change trajectories from initial to terminal states.
- Capturing dynamic changes in epigenetic states: Chromatin accessibility patterns undergo systematic changes during differentiation. Through pseudotime analysis, the open/closed states of key regulatory elements (peaks) during differentiation can be identified, revealing the dynamic patterns of epigenetic remodeling.
- Understanding cell fate determination mechanisms: The open/closed states of key regulatory elements (peaks) are closely related to cell fate determination. Pseudotime analysis helps locate fate divergence points and identify epigenetic regulatory factors driving cell fate selection.
What ATAC Pseudotime Analysis Methods Are Available? How Should I Choose?
| Tool | Core Principle | Advantages | Disadvantages | Applicable Scenarios and Recommendations |
|---|---|---|---|---|
| ATAC_Monocle3 | Pseudotime + UMAP embedding. Based on the Monocle3 framework, for the extreme sparsity of scATAC-seq data, uses LSI (Latent Semantic Indexing) preprocessing, learns graph structures on UMAP dimensionality-reduced space to infer trajectories, and calculates pseudotime values based on specified starting cells. | - Fast speed, large scale: Friendly to ultra-large ATAC datasets, capable of processing millions of cells - Epigenetic specificity: Specifically optimized for chromatin accessibility data, uses LSI dimensionality reduction to handle data sparsity - Good compatibility: Good compatibility with mainstream single-cell ATAC analysis workflows (such as Seurat) - Clear trajectories: Can identify complex branching trajectories, revealing cell differentiation paths | - Starting point dependency: Requires users to specify trajectory starting points (root cells), starting point selection greatly affects results - High data quality requirements: Requires high-quality scATAC-seq data and accurate UMAP dimensionality reduction results - Still evolving: The algorithm is undergoing rapid iteration, and its ability to resolve extremely complex branches is sometimes not as stable as transcriptome Monocle 2 | First recommendation based on scATAC-seq data. When you want to reconstruct cell differentiation trajectories based on chromatin accessibility data and identify epigenetic dynamic changes, ATAC_Monocle3 is the optimized and most reliable choice. Particularly suitable for research focusing on epigenetic regulation or early developmental events. |
Selection Suggestions:
- Based on scATAC-seq data: ATAC_Monocle3 is specifically optimized for ATAC data and is recommended.
- Multi-omics integration: If you need to analyze both scRNA-seq and scATAC-seq simultaneously, you can use Monocle 3 and ATAC_Monocle3 separately, then integrate pseudotime results from both types of data to obtain a more comprehensive perspective.
ATAC Copy Number Variation (CNV) Analysis
Definition of ATAC Copy Number Variation Analysis
Copy Number Variation (CNV) is a type of genomic structural variation, referring to an increase or decrease in the copy number of DNA fragments longer than 1kb compared to the reference genome. In single-cell ATAC sequencing (scATAC-seq) data, CNV can be indirectly inferred by analyzing the number of reads in genomic regions. If the overall number of reads in a chromosomal region increases or decreases, it may indicate that copy number gain or loss has occurred in that region.
There are two main directions for using single-cell multi-omics data for CNV analysis:
- Direction 1: Based on scRNA-seq data - Using gene expression information to infer CNV, main tools are InferCNV and CopyKAT
- Direction 2: Based on scATAC-seq data - Using read counts to infer CNV, main tools include epiAneuFinder, AtaCNV, and CopyscAT
CNV analysis based on scATAC-seq data can reveal genomic structural variations in tumor cells from an epigenetic perspective, making it particularly suitable for single-cell multi-omics research.
View More Content (Expand/Collapse)
Significance of ATAC Copy Number Variation Analysis
Single-cell ATAC CNV analysis is of great importance in tumor research:
- Distinguishing tumor cells from non-tumor cells: Tumor cells are usually accompanied by extensive genomic instability, manifested as widespread CNV. By analyzing the CNV patterns of single cells, malignant tumor cells can be effectively distinguished from normal cells mixed in tumor tissues (such as immune cells, stromal cells, etc.).
- Revealing tumor clonal heterogeneity: Tumor tissues are composed of subclones with different genomic characteristics. Single-cell CNV analysis can identify tumor subclones with different CNV patterns, thereby revealing the clonal structure and evolutionary relationships within tumors.
- Identifying key driver genes in tumors: Through CNV analysis, chromosomal regions that frequently undergo copy number gain or loss can be located. These regions may contain key oncogenes or tumor suppressor genes, providing clues for finding new therapeutic targets.
- Evaluating tumor evolution and drug resistance mechanisms: By comparing CNV patterns of tumor cells at different treatment stages or in different metastatic foci, tumor evolutionary paths can be tracked, and genomic variations related to drug resistance can be studied.
- Integrating epigenetic and genomic information: Combining chromatin accessibility data and CNV information can reveal how copy number variations affect chromatin open states, providing deeper understanding of epigenetic regulatory mechanisms.
TIP
ATAC CNV analysis is particularly suitable for studying the following biological questions:
- Malignant cell identification in tumor tissue samples
- Clonal heterogeneity and subclone structure analysis within tumors
- Early genomic structural variation detection in precancerous lesions or developmental abnormal samples
- Relationships between epigenetic states and genomic structural variations
What ATAC Copy Number Variation Analysis Methods Are Available? How Should I Choose?
| Tool | Core Principle | Advantages | Disadvantages | Applicable Scenarios and Recommendations |
|---|---|---|---|---|
| epiAneuFinder | Read counting + LSI dimensionality reduction. Uses the number of reads aligned to genomic regions in scATAC-seq data as a proxy indicator for DNA copy number, employs LSI (Latent Semantic Indexing) preprocessing, and infers CNV through statistical model analysis of read counts in each window. | - Mature algorithm: Algorithm is validated with stable and reliable results - Complete data preprocessing: Automatically removes ENCODE blacklist regions, filters low coverage windows - Adjustable windows: Default 100kb windows, can be adjusted according to data characteristics - Intuitive results: Provides clear CNV heatmap visualization | - Requires user-specified parameters: Parameters such as window size need to be adjusted according to data characteristics - Long computation time: Long computation time for large-scale data - Reference dependency: Requires sufficient cell numbers and appropriate sample types | Classic and reliable choice. When you want to use validated, stable methods for ATAC CNV analysis, epiAneuFinder is a reliable choice. Particularly suitable for standardized tumor sample analysis. |
| CopyscAT | Automatic control identification + multi-level detection. Uses read counts in scATAC-seq data to infer CNV, can automatically identify non-tumor cells as controls, and detects local CNV, fragment-level CNV, and chromosomal arm-level CNV. | - Automatic control identification: Can automatically identify non-tumor cells and use them as controls without manual specification - Multi-level detection: Can detect local CNV (such as ecDNA), fragment-level and chromosomal arm-level CNV - Double minute detection: Can identify high-copy number amplification regions in the genome - Suitable for complex tumors: Particularly suitable for handling complex samples with high levels of intratumoral heterogeneity | - Complex algorithm: Consensus clustering-based analysis method with high computational complexity - Many parameters: Requires setting multiple parameters (such as analysis method, NMF component number, etc.) - Data requirements: High data quality requirements | First choice for complex tumors. When you need to handle complex tumors with high heterogeneity (such as glioblastoma), or need to detect local CNV and double minutes, CopyscAT is the best choice. |
| AtaCNV | High-resolution detection + multi-mode normalization. Performs high-resolution CNV detection with 1 Mbp genomic fragments as units, provides four normalization modes, and improves detection accuracy through methods such as data smoothing, joint segmentation, and Bayesian inference. | - High resolution: 1 Mbp windows provide high-resolution CNV detection, capable of precisely quantifying copy number states - Multiple normalization modes: Provides four normalization modes (matching normal samples, known normal cells, all cells, automatic identification), flexibly adapting to different data situations - Data smoothing and bias removal: Automatically performs data smoothing processing and eliminates potential biases that may be caused by GC content - Automatic malignant cell identification: Automatically identifies malignant cells through CNV burden scores - Joint segmentation algorithm: Applies multi-sample BIC-seq algorithm to improve detection accuracy | - Normalization mode selection is important: Requires selecting appropriate normalization modes according to data situations, improper selection may affect results - High computational resource requirements: High-resolution detection and joint segmentation algorithms require high computational resources - High data quality requirements: Requires high-quality scATAC-seq data | First choice for high-resolution detection. When you need high-resolution CNV detection, or have matched normal samples or known normal cells, AtaCNV is the best choice. Particularly suitable for research requiring precise copy number quantification. |
Selection Suggestions:
- Standard tumor sample analysis: epiAneuFinder (mature and stable algorithm, suitable for most standardized analysis scenarios)
- Complex tumors or requiring multi-level detection: CopyscAT (automatic control identification, capable of detecting local CNV and double minutes, suitable for complex heterogeneous tumors)
- High-resolution detection or with normal references: AtaCNV (1 Mbp high resolution, multiple normalization modes, suitable for research requiring precise quantification)
- Exploratory analysis: It is recommended to first use epiAneuFinder or AtaCNV for preliminary analysis, understand data characteristics, and then decide whether CopyscAT's advanced features are needed
- Validation analysis: Multiple tools can be used simultaneously for cross-validation to ensure result reliability
- Integration with transcriptome CNV analysis: If data contains scRNA-seq information, InferCNV or CopyKAT can be used simultaneously for analysis, with results from both methods mutually validated to obtain a more comprehensive perspective
Spatial Niche Analysis
Definition
Spatial Niche Analysis identifies, characterizes, and compares spatial domains or cell niches in spatial transcriptomics. These domains consist of cell populations with specific mixing patterns, reflecting functional organization.
View More (Expand/Collapse)
Significance
- Reveal Organization: Understand functional partitioning.
- Discover Interactions: Cell mixing reflects interaction.
- Understand Pathology: Reveal disease-associated spatial remodeling.
- Guide Therapy: Provide spatial info for precision medicine.
Methods and Selection Guide
| Tool | Core Principle | Pros | Cons | Recommended Scenario |
|---|---|---|---|---|
| Banksy | Spatial Neighborhood Feature Fusion. Fuses cell expression with neighbor features for clustering. | - Integrates Spatial: Balances expression and space - Adjustable: Lambda parameter - Mature: Based on Leiden/Louvain - Rich Vis: Spatial/UMAP plots | - Parameter Sensitive: Needs tuning - Complexity: Slow on large data | Top Recommendation. Best for understanding spatial organization balancing expression and location. |
| CellCharter | Multi-sample Domain ID. Uses scVI/scArches and GMM for clustering, supports multi-sample. | - Multi-sample: Handles batch effects - Scalable: Million cells - Multi-modal: Supports multi-omics - Auto Opt: Finds best k - GPU: Fast | - Complex: Deep learning based - Setup: Needs config - Interpretation: Needs bio knowledge | Efficient for multi-sample or massive datasets. |
Selection Advice:
- Standard: Banksy.
- Multi-sample/Large: CellCharter.
Spatial Cell-Cell Communication Analysis
Definition
Infers L-R signaling integrating spatial location. Unlike traditional analysis, it considers proximity, distance, and diffusion to filter false positives and infer direction/strength.
View More (Expand/Collapse)
Significance
- Reveal Real Interactions: Filter distant pairs.
- Infer Direction: Identify sender/receiver (e.g., COMMOT).
- Understand Microenvironment: Reveal specific patterns in tumor/immune areas.
- Discover Diffusion: Capture long-range signals.
Methods and Selection Guide
| Tool | Core Principle | Pros | Cons | Recommended Scenario |
|---|---|---|---|---|
| COMMOT | Optimal Transport. Models ligand distribution to infer direction and downstream response. | - Direction: Infers flow - Diffusion: Models spread - Downstream: Predicts targets | - Complex: Slow - Params: Signal type - Data: Needs coordinates | Directional Signal Study. Best for development/morphogenesis. |
| CellPhoneDB_spatial | Spatially Constrained Permutation. Tests significance within spatial microenvironments. | - Authoritative DB: High confidence - Complexes: Handles subunits - Strict Stats: Reliable - Microenv: Compares domains | - No Direction: L-R only - Dependent: Needs domains - Human: Mostly human | Strict Validation. Best for high-confidence L-R ID. |
| CellChat_spatial | Spatially Constrained Mass Action. Adjusts probability by distance. | - Pathway: System view - Roles: Sender/Receiver - Rich Vis: Many plots - Flexible: Distance factor | - No Direction: At cell type level - Params: Many - Slow: Large data | System Pathway Analysis. Best for network comparison. |
| stLearn | SCTP + SME. Integrates space, morphology, expression. | - Low False Positive: Dual permutation - SME: Biological reality - Specific: Proximity only | - Single Sample: One at a time - Visium: Optimized for spots | Visium Validation. Best for spot-based data. |
| NicheNet_spatial | Target Gene Prediction. Integrates prior knowledge to predict how ligands affect target gene expression in receiver cells. | - Downstream Focus: Links ligand to target activation - Activity Ranking: Scores ligands - Regulatory Axis: L-R-Target network | - Prior Dependent: Limited by database - Target Sensitive: Depends on input genes - Unidirectional: Sender to receiver | Microenvironment Targets. Best for finding ligands driving specific target genes. |
Selection Advice:
- Direction: COMMOT.
- Strict Validation: CellPhoneDB_spatial.
- System View: CellChat_spatial.
- Microenvironment Targets: NicheNet_spatial.
- Visium: stLearn.
- Exploratory Analysis: Start with CellChat_spatial, use COMMOT for directionality, or NicheNet_spatial for upstream driving ligands.
Spatial Co-localization Analysis
Definition
Quantifies spatial proximity of biological entities (cell types or molecules). Identifies aggregation (positive) or repulsion (negative) to provide spatial evidence for interactions.
View More (Expand/Collapse)
Significance
- Parse Microenvironment: Reveal functional zones.
- Identify Interactions: Proximity implies potential interaction.
- Assess Differences: Compare conditions.
Methods and Selection Guide
| Tool | Core Principle | Pros | Cons | Recommended Scenario |
|---|---|---|---|---|
| MISTy | Multi-view Modeling. Models feature variation using intra/intercellular views. | - Multi-view: Reveals complex regulation - Cell Type Level: Assesses impact - Scalable: Multi-modal | - Annotation: Needs high quality - Slow: Training takes time - Complex: Interpretation needed | Cell Type/Module Modeling. Best for quantifying spatial impact. |
| SpaGene | L-R Co-localization. Identifies proximate L-R pairs via statistics. | - Focus: L-R pairs - Strict: Permutation test - Comparative: Across groups | - Threshold: Sensitive to expression - Distance: Sensitive to radius - Needs Info: Spatial evidence only | L-R Co-localization. Best for screening signal pairs. |
Selection Advice:
- Interaction Strength: MISTy.
- L-R Pairs: SpaGene.
References
- Tavernari D. et al. MISTy: Multiview intercellular spatial modeling for discovering cell neighborhoods. Nature Communications 14, 1747 (2023).
- Zhang L. et al. SpaGene identifies spatially proximate ligand–receptor interactions in tissues. Nature Communications 14, 5679 (2023).
- Armingol E. et al. Deciphering cell–cell interactions and communication from gene expression. Nature Reviews Genetics 22, 71–88 (2021).
Mutation Analysis
Definition
Detects SNVs and Indels from scRNA-seq data. Integrates mutation and expression matrices to locate mutant cells and assess functional changes.
lncRNA-mRNA Co-expression Analysis
Definition
Identifies expression correlation between lncRNA and mRNA to infer regulatory functions and functional modules.
View More (Expand/Collapse)
Significance
- Reveal Function: Infer lncRNA targets.
- Discover Modules: Identify co-expressed networks.
- Identify Markers: Cell-type-specific lncRNAs.
Tool: Use LncCoExpression.
Fusion Analysis
Definition
Detects gene fusion events from scRNA-seq (e.g., STAR-Fusion), identifies fusion-carrying cells, and assesses distribution and function.
Alternative Splicing Analysis
Definition
Compares splicing patterns (e.g., skipped exons) across conditions to identify regulation.