Blogs

Single-cell RNA sequencing (scRNA-seq) has transformed how we study complex tissues - revealing cell-type diversity, disease mechanisms, and gene expression at unprecedented resolution. But for many researchers, the path from raw public data to meaningful visualization remains a bottleneck. Downloading data, running preprocessing pipelines, managing software dependencies - it's a lot before you even start asking biological questions.

This is where the combination of Biomni, a general-purpose biomedical AI agent, and BxGenomics scRNA-Seq View, an interactive visualization and integrative analysis platform, opens up a compelling new workflow. Here, we walk through how we used Biomni to process two publicly available scRNA-seq datasets and imported the results directly into BxGenomics for interactive exploration - no local pipeline setup required.

The Datasets

We selected two human scRNA-seq datasets representing very different disease contexts to show the breadth of what this workflow can handle.

Dataset 1 - Multiple Sclerosis Cortex (PRJNA544731)
Published in Nature (Schirmer et al., 2019), this single-nucleus RNA-seq dataset profiles human cortical tissue from individuals with multiple sclerosis (MS) and healthy controls. Analyzing over 48,000 nuclei from 21 cortical samples, the study uncovered extensive cell-type-specific transcriptional alterations across excitatory neurons, oligodendrocytes, astrocytes, and microglia, revealing selective neuronal vulnerability and widespread glial remodeling in the MS cortex.

Dataset 2 - Choroidal Cells in Macular Degeneration (GSE183320)
According to Voigt et al., 2022, this dataset profiles over 30,000 human choroidal cells from 21 donors in a study evaluating age-related macular degeneration pathology, 11 of which had clinical and/or histological evidence of atrophic (n = 9) or neovascular (n = 2) macular degeneration, and 10 healthy controls. Generated with 10X Genomics Chromium and sequenced on Illumina NovaSeq, it captures endothelial cells, tissue-resident macrophages, inflammatory macrophages, arterioles, and venules, shedding light on early degenerative changes in the choriocapillaris.

Together, these two datasets span distinct tissues, disease mechanisms, and sequencing strategies, providing an excellent opportunity to demonstrate the flexibility and generalizability of the Biomni-to-BxGenomics workflow.

Step 1: Using Biomni to Generate h5ad Files

Rather than manually downloading raw reads, running Cell Ranger, and configuring Seurat or Scanpy environments, we turned to Biomni, a general-purpose biomedical AI agent originally developed by researchers at the Stanford SNAP Lab. Biomni is capable of autonomously executing bioinformatics tasks from natural language instructions, and a free online version is available through the Phylo platform.

We simply described each dataset to Biomni and asked it to retrieve, process, and output the data as .h5ad files - the standard AnnData format used across the single-cell ecosystem. Biomni handled data retrieval and processing, and returned ready-to-use h5ad files for each dataset. The key advantage: no local environment to configure, no dependency conflicts, and no pipeline scripting - just a natural language request and a file ready for downstream use. Fig. 1 illustrates the Biomni conversational interface, highlighting its key components: the Projects panel, chat window, to-do list, and results window.

Tip: When requesting h5ad output from Biomni, specify that the file should be generated using the latest AnnData version, as older AnnData formats may not be fully compatible with BxGenomics scRNA-Seq View.

Figure 1. Biomni conversational interface processing scRNA-Seq from natural language query to h5ad output.

Step 2: Importing h5ad Files into BxGenomics scRNA-Seq View

With the h5ad files in hand, we imported them into BxGenomics scRNA-Seq View, BioInfoRx's cloud-based platform for single-cell data analysis and visualization. The import process is straightforward: upload the h5ad file, and BxGenomics automatically parses the cell metadata, embeddings, and gene expression matrix. No software installation or configuration is required. Within minutes, both datasets are available in the BxGenomics environment (Dataset 1 and Dataset 2). Once loaded, the system unlocks a powerful suite of interactive tools for immediate secondary and tertiary single-cell data exploration. Researchers can seamlessly display and filter cell clusters using pre-computed dimensionality reductions like UMAP or t-SNE embeddings. Fig 2. displays the GSE183320 dataset in CellxGene VIP; cells are visualized using a UMAP embedding and colored by cell type.

Figure 2. BxGenomics scRNA-Seq View after importing choroidal cells in Macular Degeneration  (GSE183320).

Step 3: Analyzing data with Cellxgene VIP in BxGenomics scRNA-Seq View

To demonstrate the analytical capabilities of Cellxgene VIP, we examined the relative proportions of MS and control cells across major cell types in PRJNA544731. In Cellxgene VIP, users can generate a stacked bar plot by selecting cell type and diagnosis as annotations, specifying cell type for coloring, and displaying the results as proportions rather than raw counts. As shown in Fig. 3, microglia in this dataset are disproportionately derived from MS samples, consistent with the well-established role of neuroinflammation and microglial activation in multiple sclerosis pathology. 

Figure 3. Stacked bar chart showing the proportional contributions of control and MS cells across major cell-type clusters in Cellxgene VIP (PRJNA544731).

From Public Data to Biological Insight in Minutes

The Biomni-to-BxGenomics workflow significantly lowers the barrier to single-cell data exploration for both wet-lab and computational researchers. Instead of spending time downloading datasets, configuring software environments, and running preprocessing pipelines, researchers can move directly from a GEO or SRA accession number to interactive visualization and biological interpretation. Processed datasets are stored in the cloud, making it easy to share analyses with collaborators, lab members, and reviewers.

Both datasets featured in this article are now available in BxGenomics scRNA-Seq View (Dataset 1 and Dataset 2), where users can explore the processed data through an interactive web interface. Researchers can log in as guests to browse publicly available datasets, while the BioInfoRx team can also assist in uploading private h5ad files for secure interactive analysis. For users interested in automating data retrieval and preprocessing, Biomni is available online, with its source code openly accessible on GitHub.

By combining AI-driven data processing with cloud-based interactive analytics, Biomni and BxGenomics provide an end-to-end workflow that accelerates the path from public datasets to biological discovery, making single-cell analysis faster, simpler, and more accessible than ever before.