Zubair Khalid

Virologist/Molecular Biologist | Veterinarian | Bioinformatician

Conventional & Molecular Virology • Vaccine Development • Computational Biology

Dr. Zubair Khalid is a veterinarian and virologist specializing in conventional and molecular virology, vaccine development, and computational biology. Dedicated to advancing animal health through innovative research and multi-omics approaches.

Dr. Zubair Khalid - Veterinarian, Virologist, and Vaccine Development Researcher specializing in Computational Biology, Multi-omics, Animal Health, and Infectious Disease Research

Category: Guides

Cell Genomics

Cell genomics is the high resolution analysis of genomes, transcriptomes, and epigenomes at the level of individual cells. This guide is for bench scientists, bioinformatics beginners, and graduate students who want a grounded, source bounded framework for designing single cell experiments, analyzing data, and interpreting results. The NCBI Bookshelf offers essential background on genomic technologies that underpin this field NCBI Bookshelf. EMBL EBI Training provides modular courses on single cell data processing EMBL EBI Training.

At a Glance

Concept Description Key Consideration
Single cell RNA seq Measures gene expression in thousands of individual cells High dropout rate requires careful normalization
Single cell ATAC seq Maps chromatin accessibility per cell Requires high cell number for robust peak calling
Single cell DNA seq Detects somatic mutations and copy number variants Low coverage per cell limits variant detection
Multi omics integration Combines transcriptome, epigenome, and proteome from same cell Computational alignment is computationally intensive
Data analysis workflow From raw reads to clustering and differential testing Quality control steps are non negotiable

Core Concepts and Decision Criteria

Single cell genomics captures heterogeneity that bulk sequencing averages away. The main technologies each address a different biological layer. Single cell RNA seq (scRNA seq) quantifies gene expression. Single cell ATAC seq (scATAC seq) profiles open chromatin. Single cell DNA seq identifies genomic alterations. Multi omics approaches like the one used in a recent lung adenocarcinoma study combine super enhancer signatures with machine learning to identify prognostic biomarkers Multi omics machine learning investigation.

Your choice depends on your question. If you need to characterize cell types and states, scRNA seq is the standard first step. If you are studying regulatory elements, scATAC seq is more direct. For lineage tracing or mutation detection, single cell DNA seq is appropriate, though it suffers from allele dropout. A recent study of Shank3b deficiency used single cell approaches to link GABAergic inhibition deficits to anxiety like behaviors Shank3b deficiency anxiety. That example shows how cell genomics can pinpoint cell type specific mechanisms.

Decision criteria include:

  • Cell type complexity: More diverse samples benefit from higher cell counts.
  • Required resolution: Epigenomic marks need deeper sequencing per cell than transcriptome.
  • Budget and throughput: Droplet based methods (e.g., 10x Genomics) are cost effective for thousands of cells, while plate based methods give fuller coverage per cell but lower throughput.
  • Multi omics need: If you need matched transcriptome and epigenome from the same cell, use commercial kits (e.g., 10x Multiome) or custom approaches.

For a practical starting point, the Galaxy Training Network has workflows for preprocessing scRNA seq data Galaxy Training Network.

A Practical Workflow for Cell Genomics

Implementation follows a sequence of wet lab and computational steps.

  1. Sample preparation. Dissociate tissue into a single cell suspension while preserving viability above 80%. Dead cells introduce cytoplasmic RNA that contaminates libraries. For frozen samples, use a validated protocol to avoid RNA degradation.

  2. Library construction. Choose a platform. Droplet based methods (10x Genomics, Drop seq) are most common. For scATAC seq, use a transposase based kit. Always include a positive control (e.g., a mix of cell lines) and a negative control (water or empty droplets).

  3. Sequencing. Follow recommended read depths. For scRNA seq, target 20,000 to 50,000 reads per cell. For scATAC seq, target 10,000 to 50,000 fragments per cell. Sequence on an Illumina platform, the Sequence Read Archive stores raw data from published studies for reuse NCBI Sequence Read Archive.

  4. Preprocessing. Demultiplex, align reads to a reference genome, and generate a gene expression matrix. Use Cell Ranger (10x) or software from Bioconductor like DropletUtils for droplet based data Bioconductor. Filter empty droplets with ambient RNA correction (e.g., emptyDrops).

  5. Quality control. Remove cells with very few genes (likely debris), very high mitochondrial content (stressed cells), and doublets (detected by DoubletFinder or similar). Plot QC metrics to set thresholds.

  6. Normalization and batch correction. Use SCTransform or scran for normalization. Use Harmony, Seurat's IntegrateData, or scVI for batch correction when combining multiple samples. Avoid overcorrecting that removes biological variation.

  7. Dimensionality reduction and clustering. Run PCA on highly variable genes, then UMAP or t SNE for visualization. Cluster with a graph based method (Louvain or Leiden). Validate cluster stability with bootstrapping.

  8. Differential expression and annotation. Find marker genes for each cluster and assign cell types using curated references (e.g., the Human Cell Atlas). For scATAC seq, use peak to gene links. A recent study of B lineage bifurcation used epigenetic signatures to map early transcriptional divergence Epigenetic signatures early B lineage.

  9. Downstream analysis. Perform trajectory inference (Monocle, Slingshot), RNA velocity, or gene regulatory network inference. For multi omics data, use integrative tools like MOFA or CiteFuse. The Triticeae auxin response factor study demonstrates how cell genomics can be applied to plant stress responses Evolution trajectory auxin response.

Common Mistakes and How to Avoid Them

  • Insufficient cell viability. Viability below 80% increases ambient RNA and cell debris. Measure viability with trypan blue or automated counter immediately after dissociation. Use a debris removal step (e.g., density gradient or magnetic beads) if needed.

  • Batch effects ignored. Processing samples on different days or by different operators creates non biological variation. Plan to randomize batches and include a technical replicate of a reference cell line across batches.

  • Overclustering. Setting the clustering resolution too high produces many small clusters that may not be reproducible. Use silhouette score or clustree to evaluate cluster stability. Markers should be biologically plausible.

  • Using raw counts without normalization. Directly comparing raw counts across cells is invalid because of library size differences. Always normalize and log transform or use a model based method.

  • Assuming clusters are cell types. Clusters can reflect states, cell cycle phase, or stress responses. Validate cell type identity with independent markers or orthogonal methods like flow cytometry.

  • Ignoring dropouts. Single cell data has many zeros. Do not interpret a zero as absence of expression. Use statistical models that account for dropout (e.g., SCE from Bioconductor).

Limits and Uncertainty

Cell genomics has technical and biological boundaries. Technical noise includes low capture efficiency, amplification bias, and batch effects. Dropout rates can exceed 50% for lowly expressed genes. This limits detection of rare transcripts. The EMBL EBI Training resource on single cell RNA seq covers these issues in detail.

Biological uncertainty arises from cell state plasticity. A snapshot from dissociation may not reflect in vivo conditions because the dissociation process itself alters gene expression. Interpret cell type annotations as provisional.

Data integration across modalities remains challenging. Multi omics single cell data (e.g., RNA + ATAC) requires careful alignment of latent spaces. The giant virus structural interactomics study shows an alternative approach at the protein level, but cell genomics typically requires integrating distinct molecular layers Integrative structural interactomics giant virus. Small sample sizes also lead to false discovery. Use multiple biological replicates and apply multiple testing correction.

Interpretation boundaries: single cell genomics correlates features but does not establish causation. For causal inference, follow up with perturbations (CRISPR, drug treatments). The SERENA 6 trial example shows how genomic markers (ESR1 mutations) can guide treatment switching, but that application is clinical, not basic discovery Switching to camizestrant ESR1.

Frequently Asked Questions

Q1: How many cells do I need for a scRNA seq experiment? For a typical tissue with 5-10 major cell types, aim for at least 3,000 to 10,000 cells per sample to capture rare populations. More cells increase power for differential expression but also cost. Pilot with lower numbers to verify protocol.

Q2: Can I combine scRNA seq and scATAC seq data from different cells? Yes, but integration requires computational methods like Seurat's weighted nearest neighbor or the MOFA package. Better to use a multi omics kit that profiles both from the same cell for matched data.

Q3: How do I know if my clustering is biologically meaningful? Clusters should have stable marker genes that match known cell type signatures. Use the SingleR package or the Human Cell Atlas reference to annotate. If clusters lack markers, lower the resolution or check for technical artifacts.

Q4: What is the best way to share my processed data? Provide the count matrix, cluster annotations, and raw FASTQ files to public repositories. The Sequence Read Archive stores raw sequencing data. For processed data, use GEO or a dedicated single cell database like CELLxGENE.

References and Further Reading

Related Articles