Biotechnology · Advanced

Statistical and Bioinformatics Methods for Analysis of Single-Cell RNA-Seq Data

Written by SignalsLab

Early edition: these lessons are being reviewed.

Content

Summary

Single-cell RNA sequencing (scRNA-seq) profiles individual cells to capture intra-tumor heterogeneity, requiring specialized computational pipelines distinct from bulk RNA-seq due to high technical noise and zero inflation. The analysis workflow involves rigorous quality control to filter low-quality cells and doublets, followed by normalization methods that preserve biological heterogeneity while correcting for technical batch effects.

Intuition

Think of scRNA-seq as taking a high-resolution portrait of a tumor, whereas bulk RNA-seq is like a blurry group photo where everyone's features are averaged out. To get a clear picture, we must first clean the image by removing 'bad pixels' (low-quality cells or doublets) using metrics like mitochondrial transcript percentages. Then, we must adjust the lighting (normalization) carefully; if we use a standard bulk adjustment, we might distort the unique features of individual cells. The goal is to keep the true biological differences between cells visible while removing the artificial noise introduced by the sequencing process, allowing us to identify the most informative genes (HVGs) that define cell types.

Derivation

  1. scRNA-seq quantifies the transcriptome of individual cells to capture intra-tumor heterogeneity at unprecedented resolution compared to bulk RNA-seq.
  2. Droplet-based platforms encapsulate cells with unique molecular identifiers (UMIs) and barcodes to diminish sequencing read biases caused by library amplification.
  3. Quality control identifies low-quality cells using metrics like transcript count per droplet (indicating doublets or poor capture) and mitochondrial transcript percentage (indicating stress).
  4. Gene-level QC excludes genes expressed in only a small proportion of cells to reduce technical noise.
  5. Normalization adjusts for batch effects and technical noise, which are prominent in scRNA-seq due to zero inflation and cell-specific variability.
  6. Global-scaling normalization methods from bulk RNA-seq are insufficient because they cannot accommodate cell-specific variability, potentially leading to biased estimation of scaling factors.
  7. Highly variable genes (HVGs) are selected to distinguish cell populations, a process heavily influenced by the normalization method used.

Worked example

Problem

A researcher analyzes a scRNA-seq dataset comparing tumor and normal cells. They apply a global-scaling normalization method inherited from bulk RNA-seq and set stringent QC cutoffs for mitochondrial percentage. What are the potential risks?

  1. Global-scaling methods assume uniform variability, but scRNA-seq has high cell-specific variability, leading to biased scaling factors.
  2. Stringent QC cutoffs may remove rare cell populations of interest, particularly when comparing tumor to normal cells.
  3. The normalization choice affects the identification of highly variable genes (HVGs), potentially masking true biological heterogeneity.

Answer The researcher risks biased estimation of scaling factors due to cell-specific variability and the inadvertent removal of rare cell populations, which compromises the accurate identification of biologically relevant HVGs.

Common mistakes

  • Assuming global-scaling normalization methods from bulk RNA-seq are sufficient for scRNA-seq, which can lead to biased estimation of scaling factors due to cell-specific variability.
  • Applying overly stringent QC cutoffs without considering cell composition, which can inadvertently remove rare cell populations of interest.
  • Ignoring the impact of normalization on the selection of highly variable genes (HVGs), as the choice of HVGs is highly affected by the normalization method.

Animation

Animation coming soon

The animated version of this lesson is being prepared for the web. In the Signals AI app, lessons are drawn step by step.

Code

This is a text lesson: it has no animation code yet.

ccBYNC© 2026 SignalsLab. Lesson text shared as is under CC BY-NC 4.0.