No-Code Marker Gene Identification Workflow in ScarfWeb

How can researchers perform marker gene identification without writing code? ScarfWeb streamlines the single-cell analysis workflow into a series of interactive steps. For background on marker types, statistical parameters, and validation, see our guide to marker gene detection in scRNA-seq. Here is a step-by-step no-code workflow for identifying marker genes in ScarfWeb.

1. Data Upload and Setup

Figure 1: Upload datasets and assign metadata for downstream multi-condition comparisons.

Upload your single-cell data to the platform. ScarfWeb supports input of a raw count matrix or processed data from common pipelines (10x, BD, Seurat, etc.). If you have multiple samples or conditions, you can upload and assign metadata (e.g., condition labels) for each cell. This sets the stage for multi-condition comparisons later.

2. Quality Control & Preprocessing

Through the web interface, perform quality control filters (e.g., remove cells with low reads or high mitochondrial gene percentage) using simple sliders or preset options. Next, initiate normalization and scaling of the data. ScarfWeb automatically handles library size normalization and log transformation. If multiple datasets or batches are present, enable the batch correction/integration option. ScarfWeb will correct batch effects so that downstream marker analysis is not confounded by technical differences. See our batch effect correction guide or the data integration KB article. For a broader platform comparison, read our scRNA-seq analysis tools guide for 2026.

Figure 2a: Interactive quality control interface with adjustable parameters for cell filtering and data normalization.

Figure 2b: Batch correction settings panel for removing technical variation between datasets prior to analysis.

3. Clustering and Embedding

Run clustering analysis by selecting the desired resolution (a slider for coarse to fine clustering) and choose a dimensionality reduction method (PCA for analysis, UMAP/t-SNE for visualization). The platform computes principal components, lets you choose how many to use, and then clusters the cells (often via graph-based methods like Leiden).

In ScarfWeb, this might be as easy as clicking "Start Analysis" (default parameters help users get started) and waiting for the results to appear as colored clusters on a UMAP plot. No coding needed. Just interpret the resulting clusters.

Figure 3: Dimensionality reduction and clustering interface with adjustable parameters for UMAP visualization and cluster resolution.

4. Marker Gene Identification (Cluster Markers)

Once clusters are defined, use the Find Markers function. In ScarfWeb, a default one vs all marker search runs for Leiden clusters when you start an analysis. The platform performs differential expression under the hood and returns a table of marker genes for each cluster, complete with marker scores, mean expression, and fraction expressed. Here you can interactively sort genes in the table by score, fold change, etc., and search for specific genes within the table. The marker lists update instantly, showing you the refined selection. You can sort genes by marker score to see the top markers. Visualizations like violin plots and dot plots of the marker genes can usually be generated with one click to help you evaluate marker quality. For more detail on interpreting these markers, see the marker interpretation knowledge base article.

Figure 4a: Marker gene table displaying marker scores, mean expression, and fraction of expressing cells for each identified cluster.

Figure 4b: Cell sets interface allowing users to define custom cell populations for targeted marker gene analysis.

ScarfWeb's AI-powered gene information feature provides instant access to comprehensive gene summaries directly within the analysis workflow. When exploring marker genes, users can click on any gene of interest to receive a detailed summary about its biological function, expression patterns, and relevance to specific cellular processes or diseases. This integration pulls information from curated scientific literature and databases, automatically citing sources for accuracy and credibility. Researchers can contextualize findings without leaving the platform or searching through external resources.

Figure 4c: AI-powered gene information panel providing instant biological context for selected marker genes with automatically cited literature sources.

5. Dynamic Marker Discovery (Trajectory or Time Course)

Figure 5: Pseudotime analysis interface showing ordering of cells along a trajectory, with automatically calculated marker genes for each module displayed in the heatmap.

If your data involves a continuous process (like a time series or differentiation experiment), you can perform trajectory analysis without coding. ScarfWeb offers a pseudotime analysis module where you might select a set of cells or clusters to order along a trajectory. By clicking "Run Trajectory Analysis," the platform infers a pseudotime ordering (displayed on UMAP). When you run pseudotime analysis, marker genes in each module are calculated automatically with no extra step required. These could be early vs late markers of the process. The results are shown as a heatmap with modules, and you can visualize each gene's expression along the trajectory. This no-code approach replaces what would otherwise require coding with Monocle or tradeSeq. For more details, see the pseudotime analysis knowledge base article.

6. Multi-Condition Marker Comparison

One powerful aspect of ScarfWeb is analyzing multiple conditions side by side. If you uploaded data from, say, a control vs treated sample or healthy vs disease, you can use the interface to compare marker genes across conditions. For example, after clustering, you might identify cluster 5 as a cell type present in both conditions. ScarfWeb can find markers of cluster 5 within each condition and highlight differences. Alternatively, you can perform a differential expression between the same cell type under two conditions (to find condition-specific markers). This might involve selecting condition labels in a dropdown and running a built-in DE analysis. ScarfWeb's cell filtering feature allows users to define cell populations with custom filters and use them for deeper analysis by filtering by category, cluster, imported groups, and more. The no-code platform will output genes that are markers in one condition's cluster vs the other (an interaction effect). This is useful for finding, for instance, activation markers that only appear in disease. All without writing a line of R code!

Figure 6: Advanced cell filtering interface allowing multi-condition marker comparison through custom cell set creation based on clusters, metadata, and experimental conditions.

7. Marker Scoring and Custom Gene Sets

As you identify markers, ScarfWeb provides the Marker Score to help judge each gene. You might focus on genes with high scores (close to 1.0) as the best markers. The platform also allows you to create custom gene sets or signatures. For example, if you have a candidate list of markers from literature, you can input them and ScarfWeb will compute a signature score per cell to see if that signature aligns with your clusters. This helps validate if certain cell clusters in your analysis match known biology. Conversely, you can take your discovered markers and immediately test them: e.g., check their expression in an integrated public dataset or across another condition by loading another dataset in.

Figure 7: Custom gene set creation interface allowing users to input and analyze specific gene signatures across their single-cell data.

8. Export and Annotation

Finally, with marker genes identified, a no-code platform lets you easily export the results. You can download the marker gene tables as CSV files directly. ScarfWeb lets you share the dataset with collaborators through its cloud interface. At this stage, you would use the marker genes to annotate clusters. The platform has an annotation tool where you can label cluster IDs with cell type names manually, or use AI-powered auto-annotation (Nygen Insights auto-annotation applies known marker databases to suggest cell types with confidence scores). For a broader annotation workflow, see our scRNA-seq cluster annotation guide. By the end of this workflow, you have identified and labeled all your cell populations with marker genes, all through a point-and-click experience.

Figure 8: Nygen Insights automated cell annotation feature, which applies machine learning to identify cell types based on gene expression patterns and provides confidence scores for each annotation.

This no-code workflow mirrors the steps an expert would do via programming, but it lowers the barrier to entry. It also speeds up the iteration cycle. You can tweak parameters and re-run analyses in minutes. The result is that even complex analyses like multi-condition marker discovery or trajectory-based gene detection become accessible to researchers who prefer not to code.