Can a 91.8% RNA-seq Mapping Rate Tell the Whole Story? Bioinformatics Project

Ambuj Narayan Maurya — Sun, 25 Oct 2026 09:55:52 +0000

Hey DEV Community! 👋

This is my first post here, and I'm excited to connect with developers, bioinformaticians, and researchers who enjoy solving scientific problems with code.

I'm building my career in bioinformatics, and one question that came up during my recent project was:

If most sequencing reads map to the reference genome, does that automatically mean the analysis is reliable?

My zebrafish RNA-seq project taught me why the answer is more complicated.

The project

I built a workflow to analyze paired-end RNA sequencing data from two developmental stages of zebrafish (Danio rerio): the 2-cell stage and the 6-hour stage.

I used Nextflow DSL2 to organize the analysis into a single workflow.

The tools included:

FastQC and MultiQC:Sequencing quality control
TopHat:Splice-aware read alignment
Cufflinks:Transcript assembly and expression estimation
Cuffdiff:Differential expression analysis
Custom processing:Extraction of candidate gene identifiers

The workflow completed successfully, with nine tasks finishing without failure.

What did the alignment results show?

Here are two metrics from the analysis:

Metric 2-cell stage 6-hour stage

Overall read mapping rate 91.8% 89.5%
Concordant pair alignment rate 82.8% 75.4%

The overall mapping rates differed by 2.3 percentage points. However, the concordant pair alignment rates differed by 7.4 percentage points.

Why are these metrics different?

Overall mapping rate tells us how many reads align to the reference genome.

Concordant pair alignment assesses whether paired reads align in the expected relationship and orientation.

This comparison showed me why a single mapping percentage cannot describe every aspect of alignment quality.

The lower concordant pair alignment rate in the 6-hour sample warrants further investigation. However, these metrics alone cannot explain the cause or establish a biological effect.

What about differential expression?

The analysis produced four gene identifiers passing the configured Cuffdiff q-value threshold of 0.01.

But there is an important limitation: the dataset contains only one sample per developmental stage.

Without biological replicates, biological variability cannot be estimated reliably. Therefore, these gene identifiers should be treated as exploratory candidates, not validated differential expression findings.

This was one of the most important lessons from the project.

A pipeline can finish successfully without providing enough evidence for a strong biological conclusion.

What I learned

This project helped me connect several areas of bioinformatics:

Sequencing quality control and alignment interpretation
Transcript assembly and expression estimation
Workflow automation using Nextflow
Software version tracking and reproducibility
Scientific interpretation and experimental design limitations

I also documented the workflow, software environment, results, and limitations in my GitHub repository.

What's next?

I want to explore modern RNA-seq workflows using maintained alignment tools and statistically appropriate methods for replicated experiments.

My goal is to understand not just how to run bioinformatics tools, but also why each step matters and what the results can legitimately tell us.

Explore the code and documentation: Zebrafish RNA-seq Workflow on GitHub ( https://github.com/Ambuj-NMomics/RNA-seq-analysis-on-Zebrafish)

Since this is my first post on DEV, I'd love to connect with people working in bioinformatics, genomics, computational biology, and scientific programming.

What's one lesson your first real data analysis project taught you?

Thanks for reading! 👋

DEV Community: Ambuj Narayan Maurya

Can a 91.8% RNA-seq Mapping Rate Tell the Whole Story? Bioinformatics Project