After evaluating raw sequencing quality, the next stage of the RNA-Seq system is to transform sequencing reads into quantitative measurements of gene expression.
At this point, the goal is no longer to assess data quality but to determine how many reads support each biological feature, such as a gene or transcript.
The output of this stage forms the foundation for downstream statistical analysis.
flowchart TD
A[Sequencing]
subgraph DP["Data Processing"]
B[Raw Reads]
C[Read Quality Control]
D[Read Processing & Quantification]
E[Count Matrix]
end
A --> B
B --> C --> D --> Eflowchart TD
A[Sequencing]
subgraph DP["Data Processing"]
B[Raw Reads]
C[Read Quality Control]
D[Read Processing & Quantification]
E[Count Matrix]
end
A --> B
B --> C --> D --> E
This chapter focuses on converting quality-assessed sequencing reads into quantitative expression measurements.
Quantification is the process of estimating expression levels from sequencing reads.
The central question is:
Which genes or transcripts generated these reads?
The answer allows us to summarize millions of sequencing reads into biologically meaningful expression measurements.
Most RNA-Seq workflows use a reference genome or transcriptome.
Reads are compared against known biological sequences to determine their likely origin.
Common references include:
The quality of the reference influences downstream quantification accuracy.
Traditional RNA-Seq workflows often begin with sequence alignment.
Alignment attempts to determine where each read originated in the genome.
Common aligners include:
Typical workflow:
FASTQ Files
↓
Alignment
↓
Aligned Reads (BAM)
↓
Feature Counting
↓
Count MatrixAlignment provides detailed mapping information but may require substantial computational resources.
Modern workflows often use pseudoalignment or lightweight mapping approaches.
Examples include:
Rather than performing full genomic alignment, these tools estimate transcript abundance more efficiently.
Typical workflow:
FASTQ Files
↓
Pseudoalignment
↓
Transcript Quantification
↓
Gene-Level Summarization
↓
Count MatrixThese methods are often faster while maintaining high accuracy for expression estimation.
Expression can be quantified at different biological levels.
Gene-level quantification summarizes expression across all transcripts associated with a gene.
Example:
GeneA = 250 reads
GeneB = 1020 reads
GeneC = 430 readsGene-level analyses are commonly used for differential expression studies.
Transcript-level quantification estimates abundance for individual transcript isoforms.
Example:
GeneA-001 = 150 reads
GeneA-002 = 100 readsTranscript-level analyses can provide additional biological detail but are often more complex.
For alignment-based workflows, reads are commonly assigned to annotated features.
Features may include:
Feature counting converts mapped reads into a numerical expression matrix suitable for statistical analysis.
The primary output of this stage is a count matrix.
| Gene | Sample1 | Sample2 | Sample3 |
|---|---|---|---|
| GeneA | 250 | 310 | 295 |
| GeneB | 1020 | 980 | 1105 |
| GeneC | 430 | 390 | 470 |
Rows represent biological features.
Columns represent samples.
Values represent observed counts.
This matrix becomes the foundation for downstream expression analysis.
Typical outputs include:
These outputs help evaluate how successfully reads were assigned to biological features.
Mapping rate refers to the proportion of reads successfully assigned to the reference.
For example:
Total Reads: 20,000,000
Mapped Reads: 18,500,000
Mapping Rate: 92.5%Mapping rates can help identify potential issues with:
Mapping rates should be interpreted together with other quality metrics.
Some reads may align to multiple locations.
This can occur because of:
Different quantification tools handle multi-mapping reads differently.
Understanding these decisions is important when interpreting expression estimates.
Longer genes naturally generate more reads than shorter genes.
For this reason, raw counts are influenced by:
Downstream normalization methods help account for some of these effects.
Frequently used RNA-Seq quantification tools include:
| Tool | Purpose |
|---|---|
| STAR | Alignment |
| HISAT2 | Alignment |
| featureCounts | Gene counting |
| Salmon | Quantification |
| kallisto | Quantification |
| tximport | Import and summarize transcript estimates |
The specific tool is less important than understanding the role it plays in the RNA-Seq system.
salmon quant \
-i transcriptome_index \
-l A \
-1 sample_R1.fastq.gz \
-2 sample_R2.fastq.gz \
-o sample_quantThis command estimates transcript abundance from paired-end sequencing reads.
library(tximport)
txi <- tximport(
files = quant_files,
type = "salmon"
)The imported object can then be used for downstream differential expression analysis.
Before moving to expression analysis, confirm that:
Common quantification mistakes include:
The goal is not simply to obtain counts but to understand how those counts were produced.
Read processing and quantification convert sequencing reads into expression measurements.
The count matrix produced at this stage becomes the central input for normalization, exploratory analysis, and differential expression modeling.
Understanding how counts are generated is essential for interpreting downstream biological conclusions.
The next chapter focuses on count matrix quality assessment and filtering, the first step in preparing expression measurements for statistical analysis.