Help & Documentation

Navigation

• Data summary
• 1. Overview of RESIST
  • 1.1 Data collection and processing method
  • 1.2 Cell clustering and annotation
  • 1.3 Spatial Transcriptomic Analysis
  • 1.4 Cell-Cell Communication Analysis
  • 1.5 Differential Ligand-Receptor Interaction Analysis
  • 1.6 Identification of drug resistance-associated genes
  • 1.7 Pathway enrichment analysis
  • 1.8 Intratumoral heterogeneity (ITH) and EMT scoring
  • 1.9 microRNA prediction
  • 1.10 RNA-binding protein enrichment
  • 1.11 Drug enrichment analysis
  • 1.12 Immunogenomic analysis: HLA typing and neoantigen prediction
  • 1.13 Single-cell SNV profiling and annotation
  • 1.14 Alternative polyadenylation analysis
  • 1.15 Integrated analysis module
• 2. Dataset Analysis Page for scRNA-seq data
  • 2.1 Dataset Summary (Section 1)
  • 2.2 Cell Annotations and Drug Response Distribution (Section 2)
  • 2.3 Differential Gene Expression in Tumor Cells (Section 3)
  • 2.4 Functional Enrichment Analysis (Section 4)
  • 2.5 Intratumoral Heterogeneity (ITH) Analysis (Section 5)
  • 2.6 EMT Score Comparison (Section 6)
  • 2.7 MicroRNA Regulation (Section 7)
  • 2.8 RNA-Binding Protein Enrichment (Section 8)
  • 2.9 Motif and Transcription Factor Enrichment (Section 9)
  • 2.10 Drug Enrichment Analysis (Section 10)
  • 2.11 HLA Class I Genotyping (Section 11)
  • 2.12 IPA-Derived Neoantigen Prediction (Section 12)
  • 2.13 Structural Modeling of Neoantigen-MHC Complexes (Section 13)
  • 2.14 Single-Cell Variant Landscape (Section 14)
  • 2.15 Alternative Polyadenylation (Section 15)
• 3. Dataset Analysis Page for Spatial Transcriptomics
  • 3.1 Spatial Distribution of Annotated Cell Types (Section 3)
  • 3.2 Global and Tumor-Centered Cell-Cell Interaction Networks (Section 4)
  • 3.3 Differential Ligand-Receptor Interaction Analysis (Section 5)
  • 3.4 Spatial Localization of Resistance-Associated Genes (Section 7)
• 4. Download data and contact us

1. Overview of RESIST

RESIST is a large-scale database designed to systematically characterize cancer drug resistance using therapeutic single-cell and spatial transcriptomic datasets. Drug resistance remains a major obstacle to durable cancer treatment, yet existing studies are often fragmented across individual experiments, cancer types, and regulatory layers. As a result, it is difficult to obtain an integrated view of resistance-associated transcriptional programs and their broader biological consequences.

To address this gap, RESIST was developed with three core design principles: (i) harmonization of heterogeneous therapeutic datasets under a unified processing framework, (ii) integration of multi-layer regulatory and genomic features beyond gene-level expression, and (iii) support for systematic exploration across cancer types, drug regimens, and biological contexts.

The current version of RESIST integrates 81 single-cell RNA-seq datasets and 7 spatial transcriptomics datasets curated from 48 independent studies, comprising 522 single-cell samples and 96 spatial samples derived from both human and mouse models. These datasets span 13 major cancer types and 59 distinct drug treatment regimens, providing broad coverage of chemotherapy, targeted therapy, and immunotherapy settings.

All datasets are uniformly processed to enable cross-study comparison and are accompanied by standardized analyses of resistance-associated transcriptional changes at cellular resolution. In addition to differential expression and pathway enrichment, RESIST incorporates regulatory and genomic annotations including microRNA associations, RNA-binding protein enrichment, alternative polyadenylation dynamics, single-cell SNV profiling, HLA class I genotyping, and intronic polyadenylation-derived neoantigen prediction. The integration of spatial transcriptomics further enables investigation of tumor-microenvironment organization and cell-cell communication in resistant tissues.

By unifying large-scale therapeutic single-cell and spatial data within a coherent framework, RESIST provides a comprehensive resource for exploring resistance programs across cancers and treatment contexts, facilitating hypothesis generation for candidate biomarkers and potential therapeutic vulnerabilities.

1.1 Data collection and processing method

Therapeutic datasets were systematically collected from GEO using FDA-approved drug names and drug resistance-related keywords combined with single-cell and spatial transcriptomics search terms. Only Homo sapiens and Mus musculus datasets with available drug response information were included.

scRNA-seq preprocessing

Single-cell RNA-seq data were processed using Seurat (v4.3.0). Cells with fewer than 500 detected genes or mitochondrial gene ratio greater than 10% were excluded in most datasets. Data normalization, scaling, and identification of highly variable genes were performed following the SCTransform workflow. Batch effects, when present, were corrected using Harmony. Principal component analysis was conducted for dimensionality reduction, followed by clustering and visualization using UMAP or t-SNE.

Spatial transcriptomics preprocessing

Spatial gene expression matrices and corresponding histological images were processed using Seurat. Low-quality spots were filtered based on gene count and UMI thresholds. Normalization and highly variable gene identification followed standard Seurat workflows. Spatial clustering and visualization were performed using SpatialDimPlot, and differentially expressed genes between spatial regions were identified using Seurat::FindMarkers.

1.2 Cell clustering and annotation

Cell clustering was performed using SCTransform-normalized expression profiles. Cell type annotation was assigned based on 86 curated marker genes compiled from original publications and relevant studies. UMAP embeddings display cells colored by Seurat clusters, annotated cell types, and drug response status (resistant vs sensitive).

1.3 Spatial Transcriptomic Analysis

Spatial transcriptomic datasets were processed using Seurat. Gene expression matrices were normalized and scaled following standard workflows. Spatial coordinates and histological images were integrated with expression data to enable joint transcriptomic–histological analysis.

Spatial clustering and differential expression analyses were performed to identify region-specific gene expression patterns. Drug response–associated genes were evaluated within spatial contexts by comparing resistant and sensitive samples separately.

1.4 Cell-Cell Communication Analysis

Cell–cell communication inference was performed using the CellChat framework with curated ligand–receptor interaction databases.

Normalized expression matrices were used to calculate communication probabilities between annotated cell types based on the averaged expression of ligands and receptors. Statistical significance of interactions was assessed through permutation-based testing implemented in CellChat.

Separate analyses were conducted for resistant and sensitive groups to construct condition-specific interaction networks.

1.5 Differential Ligand-Receptor Interaction Analysis

Differential ligand–receptor interactions between resistant and sensitive groups were identified by comparing communication probabilities across conditions.

Significantly differential interactions were defined based on permutation-derived p-values and changes in communication probability. Interaction strength and directionality were quantified using CellChat-derived metrics.

1.6 Identification of drug resistance-associated genes

Differential gene expression analysis between resistant and sensitive cells was performed for each cell type using Seurat::FindMarkers. Pre- and post-treatment samples were analyzed separately when applicable.

Genes with adjusted p value < 0.05 and |avg_log2FC| > 0.25 were defined as resistance-associated DEGs. Genes with |avg_log2FC| > 1 were classified as significantly resistance-associated genes. Volcano plots highlight top upregulated and downregulated genes ranked by fold change.

1.7 Pathway enrichment analysis

Functional enrichment analysis of significantly resistance-associated genes was performed using clusterProfiler. Gene Ontology Biological Process (GO BP), KEGG pathways, and HALLMARK gene sets were evaluated. Significantly enriched pathways (adjusted p < 0.05) were visualized as bar plots, separately for upregulated and downregulated genes.

1.8 Intratumoral heterogeneity (ITH) and EMT scoring

The ITH score was defined as the average Euclidean distance between each malignant cell and all other malignant cells based on the first 20 principal components. EMT scores were calculated using GSVA with the MSigDB "Epithelial-Mesenchymal Transition" gene set.

Differences between resistant and sensitive groups were assessed using Wilcoxon tests and visualized using boxplots.

1.9 microRNA prediction

miRNAs potentially regulating resistance-associated genes were predicted using miRDB. Candidates with prediction score > 80 were retained and visualized for top DEGs.

1.10 RNA-binding protein enrichment

RBP enrichment analysis was performed using hypergeometric testing based on ENCORI/starBase target gene collections. Significantly enriched RBPs (FDR < 0.05) were visualized using circle packing plots.

1.11 Drug enrichment analysis

To identify compounds potentially reversing resistance-associated transcriptional signatures, drug enrichment analysis was performed using the LINCS L1000 dataset. Kolmogorov-Smirnov-based enrichment scores were computed comparing DEG signatures with drug-induced transcriptional profiles. Compounds with FDR < 0.05 were considered significant.

1.12 Immunogenomic analysis: HLA typing and neoantigen prediction

HLA class I genotypes were inferred from scRNA-seq BAM files using OptiType after extraction of chromosome 6 reads from malignant cells. Read support and alignment profiles across exons 2 and 3 were evaluated.

Tumor-specific intronic polyadenylation (IPA) events were identified using IPAFinder. Candidate neoantigen peptides (8–11 amino acids) were generated and evaluated for binding affinity to inferred HLA alleles using NetMHCpan. Peptides with predicted percentile rank < 2% and absent from reference proteomes were retained as high-confidence candidates.

Three-dimensional structural modeling of peptide-MHC complexes was performed using AlphaFold2-Multimer to evaluate structural feasibility.

1.13 Single-cell SNV profiling and annotation

Single-nucleotide variants were called from scRNA-seq BAM files using cellSNP-lite. Variants were annotated using dbSNP and ClinVar via a custom annotation pipeline, integrating genomic coordinates, functional consequences, and clinical interpretations.

1.14 Alternative polyadenylation analysis

Alternative polyadenylation (APA) dynamics were quantified at single-cell resolution using scUTRquant. Read enrichment (RE) scores were calculated as:

RE = log₂(RD_aUTR + 1RD_cUTR + 1)

Differential APA analysis between resistant and sensitive groups was performed using Wilcoxon testing with multiple testing correction. This framework enables systematic characterization of post-transcriptional regulation across malignant and nonmalignant cell types.

1.15 Integrated analysis module

RESIST provides an integrated dataset-level analysis module based on standardized Seurat RDS objects. Users can systematically explore clustering and cell-type annotation, differential expression analysis, pathway enrichment, ITH/EMT comparison, miRNA and RNA-binding protein regulation, drug enrichment analysis, single-cell variant landscapes, HLA class I genotyping, intronic polyadenylation (IPA)-derived neoantigen prediction, and alternative polyadenylation (APA) dynamics.

All analytical workflows are implemented within a unified framework and delivered through interactive visualizations and downloadable result tables, enabling comprehensive exploration of resistance-associated transcriptional, regulatory, immunogenomic, and post-transcriptional features without requiring local computation.

Detailed implementation of variant analysis, HLA typing, neoantigen prediction, and APA quantification pipelines is available at the RESIST GitHub repository: https://github.com/QSong-github/RESIST

2. Dataset Analysis Page for scRNA-seq data

This page presents standardized multi-layer analyses for each single-cell RNA sequencing (scRNA-seq) dataset integrated in RESIST. All analyses are performed using a unified preprocessing and computational framework to ensure methodological consistency across studies. When spatial transcriptomic data are available for a dataset, spatial-specific analyses are displayed in parallel using the same standardized structure.

2.1 Dataset Summary (Section 1)

This section provides an overview of the selected dataset, including cancer type, treatment regimen, drug category, number of samples, sequencing platform, and drug response classification. Samples are categorized into resistant and sensitive groups based on reported therapeutic response or experimental design (e.g., pre-treatment versus post-treatment). This summary establishes the biological and clinical context necessary for interpreting downstream analysis.

The dataset summary also clarifies whether the study includes both pre- and post-treatment samples or represents a single timepoint comparison. For longitudinal datasets, analyses are stratified by treatment stage to preserve temporal resolution. These metadata provide the structural foundation for all subsequent resistance-associated analyses displayed on this page.

2.2 Cell Annotations and Drug Response Distribution (Section 2)

This section visualizes the cellular composition and response distribution within the dataset. Single-cell clustering is performed using a standardized Seurat workflow, and UMAP embeddings display cells colored by Seurat clusters, annotated cell types, and drug response status (resistant versus sensitive). This enables rapid assessment of cell-type heterogeneity and treatment-associated cellular shifts.

A bar chart summarizes the proportional distribution of Seurat clusters within each drug response group. Percentages sum to 100% within each response category, allowing direct comparison of relative cluster contributions between resistant and sensitive samples. These visualizations help identify response-associated compositional remodeling at the cellular level.

2.3 Differential Gene Expression in Tumor Cells (Section 3)

This section characterizes transcriptional differences between resistant and sensitive tumor cells. Differential expression analysis is performed separately for each dataset using standardized thresholds for statistical significance and fold change. Genes meeting the defined criteria are classified as resistance-associated differentially expressed genes (DEGs).

Volcano plots display genome-wide expression shifts, highlighting significantly upregulated and downregulated genes in resistant tumor cells. The top 10 most significant genes are labeled for interpretability. For datasets containing both pre- and post-treatment samples, analyses are presented separately by treatment stage. Complete DEG tables are available in the Download section for further exploration.

2.4 Functional Enrichment Analysis (Section 4)

To interpret resistance-associated transcriptional programs, significantly regulated DEGs are subjected to pathway enrichment analysis. Enrichment is performed separately for upregulated and downregulated genes in resistant tumor cells, enabling directional interpretation of biological processes.

Section 4-1: GOBP

Section 4-2: KEGG

Section 4-3: Hallmark

Bar plots summarize the top enriched Gene Ontology Biological Processes (GOBP), KEGG pathways, and Hallmark gene sets. If no pathways are displayed for a category, this indicates that no statistically significant enrichment was detected under the predefined thresholds. These analyses provide functional context for resistance-associated gene expression patterns.

2.5 Intratumoral Heterogeneity (ITH) Analysis (Section 5)

This section evaluates differences in intratumoral heterogeneity between resistant and sensitive tumor cells. ITH scores are computed based on transcriptional variability within malignant cell populations, reflecting the degree of intra-tumor diversity across individual cells.

Or, for datasets with pre and post timepoint:

Boxplots illustrate the distribution of ITH scores across response groups. Statistical comparisons are performed using a two-sided non-parametric test. For datasets containing both resistant and sensitive samples at a single treatment stage, results are shown as one comparison panel. For datasets that include both pre-treatment and post-treatment time points, comparisons are presented separately for each treatment stage, resulting in two panels to reflect stage-specific heterogeneity changes. These results provide insight into whether resistance is associated with increased or decreased cellular heterogeneity.

2.6 EMT Score Comparison (Section 6)

This section assesses epithelial-mesenchymal transition (EMT) activity in tumor cells. EMT scores are calculated using curated EMT gene signatures and quantify the extent of mesenchymal transcriptional programs within malignant cell populations.

Or, for datasets with pre and post timepoint:

Boxplots illustrate EMT score distributions across resistant and sensitive groups. Statistical significance is evaluated using a two-sided non-parametric test. For datasets containing resistant and sensitive samples at a single treatment stage, results are shown as one comparison panel. For datasets that include both pre-treatment and post-treatment time points, comparisons are presented separately for each treatment stage, resulting in two panels to reflect stage-specific EMT dynamics. This analysis evaluates whether drug resistance is associated with enhanced mesenchymal-like phenotypes across different therapeutic contexts.

2.7 MicroRNA Regulation of Resistance-Associated Genes (Section 7)

This section explores post-transcriptional regulation mediated by microRNAs. Predicted miRNA-gene interactions are obtained from curated databases, and high-confidence regulatory pairs are identified for top resistance-associated DEGs.

Treemaps visualize predicted miRNAs regulating upregulated and downregulated DEGs separately. Block size corresponds to regulatory confidence score. Only high-confidence interactions are shown. Complete miRNA-gene association tables are available for download.

2.8 RNA-Binding Protein (RBP) Enrichment Analysis (Section 8)

This module investigates potential regulation by RNA-binding proteins. Enrichment analysis is performed to identify RBPs whose known target genes overlap significantly with resistance-associated DEGs.

Bubble plots display significantly enriched RBPs for upregulated and downregulated genes separately. Bubble size and color represent statistical significance. This analysis highlights potential RNA-level regulatory mechanisms associated with drug resistance.

2.9 Motif and Transcription Factor Enrichment (Section 9)

This section identifies transcriptional regulators associated with resistance-related DEGs. Motif enrichment analysis is performed to detect overrepresented regulatory motifs near gene promoters.

Tables display enriched motifs and their corresponding transcription factors. Upregulated and downregulated genes are analyzed separately. This module provides insight into transcriptional regulatory programs driving resistance phenotypes.

2.10 Drug Enrichment Analysis (Section 10)

This section identifies candidate compounds potentially associated with drug resistance-related transcriptional states using Connectivity Map (CMap/LINCS L1000)-based enrichment analysis. Differentially expressed genes (DEGs) from tumor cells are compared against drug-induced gene expression signatures to evaluate transcriptional similarity or reversal effects.

Bar plots display the top 10 significantly enriched compounds ranked by enrichment score. A positive enrichment score indicates that the compound may reverse resistance-associated gene expression patterns (i.e., suppress resistance-upregulated genes and enhance resistance-downregulated genes), whereas a negative score suggests concordant transcriptional effects. Statistical significance is assessed through permutation testing with multiple testing correction. The complete list of significantly enriched compounds is available in the Download section.

2.11 HLA Class I Genotyping (Section 11)

This section presents HLA class I genotyping results inferred from scRNA-seq-derived BAM files. Reads mapped to HLA loci are aggregated to generate sample-level genotype predictions.

Tables summarize predicted HLA-A, HLA-B, and HLA-C alleles along with supporting read counts and confidence scores. Read depth plots illustrate alignment profiles across allele sequences, enabling evaluation of coverage distribution and genotyping robustness.

2.12 IPA-Derived Neoantigen Prediction (Section 12)

This section summarizes predicted MHC class I neoantigens derived from tumor-specific intronic polyadenylation (IPA) events. IPA events are identified in malignant cells and filtered against matched non-malignant populations to ensure tumor specificity. Candidate peptide sequences (8–11 amino acids) are generated from novel IPA-induced transcript isoforms.

The table reports gene of origin, genomic coordinates of IPA-associated terminal exons, IPA classification and usage strength (IPUI), matched HLA allele, predicted peptide sequence, and binding affinity percentile rank (%Rank) calculated using NetMHCpan. Peptides with low %Rank values indicate strong predicted binding affinity and represent high-confidence neoantigen candidates.

2.13 Structural Modeling of Neoantigen-MHC Complexes (Section 13)

This section presents three-dimensional structural models of predicted neoantigen-MHC class I complexes. For each peptide-HLA pair, structural prediction is performed using AlphaFold2-Multimer, followed by visualization in PyMOL.

The rendered structures illustrate peptide positioning within the MHC binding groove and highlight key structural elements such as α1/α2 domains and anchor residue interactions. These models provide structural context for evaluating the feasibility of antigen presentation and potential immune recognition.

2.14 Single-Cell Variant Landscape (Section 14)

This section displays somatic single-nucleotide variant (SNV) profiles derived from single-cell RNA-seq data. Variants are identified using cellSNP-lite and summarized in a cell-by-variant matrix.

Heatmaps visualize the distribution of selected variants across individual cells in resistant and sensitive groups. Each row represents a variant labeled by genomic position, and each column represents a cell. Accompanying annotation tables integrate dbSNP and ClinVar information, including gene context, functional consequence, protein changes, and clinical classification when available.

2.15 Alternative Polyadenylation (APA) Dynamics (Section 15)

This section characterizes alternative polyadenylation dynamics at single-cell resolution. PAS usage is quantified using scUTRquant, and relative expression (RE) is calculated as log₂(aUTR/cUTR), reflecting distal versus proximal polyadenylation site usage.

ECDF plots illustrate global APA shifts across cell types between resistant and sensitive groups. Violin plots display the top differentially regulated APA-associated genes per cell type, with statistical testing performed using Seurat::FindMarkers and adjusted p-values. This module reveals cell type-specific post-transcriptional regulation patterns associated with drug resistance.

Differential analysis results for relative expression (RE)-associated genes are also included in this section. Differential testing is performed between resistant and sensitive groups using RE values as quantitative features.

The table displays genes meeting the predefined significance criteria, including an adjusted p-value < 0.05 and an absolute average log₂ fold change (|avg_log2FC|) > 0.5. These genes represent significantly regulated alternative polyadenylation events associated with drug response. The complete list of tested genes is available in the Download section.

3. Dataset Analysis Page for Spatial Transcriptomics

This page presents standardized analyses for spatial transcriptomic datasets integrated in RESIST. All spatial datasets are processed using the same quality control, normalization, and differential analysis framework applied to scRNA-seq data to ensure methodological consistency.

In addition to shared transcriptomic analyses (Section 1–8), spatial datasets include dedicated modules that leverage spatial coordinates and tissue architecture information. These spatial-specific analyses are described below.

Specifically, Sections 1 and 2 correspond to the scRNA-seq dataset overview and clustering modules, maintaining identical preprocessing, annotation, and differential expression procedures. Section 6 (Differential Gene Expression) aligns with Section 3 of the scRNA-seq analysis page. Furthermore, Sections 8–12 correspond to Sections 4–8 of the scRNA-seq workflow, including pathway enrichment, ITH/EMT scoring, miRNA analysis, and RBP enrichment.

The remaining sections (Sections 3–5, 7) are spatial-specific modules that leverage tissue coordinate information and spatially resolved ligand–receptor inference, providing additional contextual insight beyond dissociated single-cell analyses.

3.1 Spatial Distribution of Annotated Cell Types (Section 3)

This module visualizes the spatial organization of annotated cell types within tissue sections under different drug response conditions. Spatial transcriptomic capture spots are mapped to their physical coordinates and overlaid onto the corresponding histological images.

By comparing resistant and sensitive samples, users can examine how tumor cells and non-malignant populations (e.g., immune or stromal compartments) are distributed within the tissue architecture. This spatial view allows assessment of microenvironmental organization and potential structural differences associated with therapeutic response.

3.2 Global and Tumor-Centered Cell-Cell Interaction Networks (Section 4)

This module presents inferred ligand–receptor–mediated communication networks separately for resistant and sensitive groups. Interaction inference is performed using curated ligand–receptor databases combined with statistical testing to identify significant signaling events between annotated cell types. Communication probability is estimated based on aggregated ligand and receptor expression across cell populations, allowing construction of condition-specific signaling networks.

3.3 Differential Ligand-Receptor Interaction Analysis (Section 5)

This module identifies ligand-receptor interactions that differ significantly between resistant and sensitive groups by directly comparing inferred communication probabilities across conditions. Differential interactions are determined using statistical testing to detect ligand–receptor pairs whose signaling strength is significantly altered between response states.

By focusing on condition-specific changes rather than absolute interaction strength, this analysis highlights signaling axes that may contribute to drug resistance-associated microenvironment remodeling. These results complement the global interaction networks by pinpointing specific ligand-receptor pairs enriched in either resistant or sensitive contexts, facilitating mechanistic interpretation of response-associated communication shifts.

3.4 Spatial Localization of Resistance-Associated Genes (Section 7)

This module maps the expression patterns of representative drug resistance-associated genes onto spatial transcriptomic coordinates within tissue sections. Normalized gene expression values are projected onto histological images to evaluate regional transcriptional enrichment in resistant and sensitive samples separately.

By integrating differential expression results with spatial localization, this analysis allows assessment of whether resistance-associated genes are confined to specific anatomical niches (e.g., tumor core, invasive margins, immune-rich regions) or broadly distributed across the tissue.

4. Download data and contact us

Please go to the download page and contact page for data access and inquiries.

Copyright 2025-Present - University of Florida

Emergency Information|Site Policies