How to Set Up a Quality Control Pipeline for Long-Read RNA-Seq Data: From Raw Reads to Isoform Quantification

By Dr. Zubair Khalid, DVM, MS, PhD · 2026-08-21

Long-read RNA sequencing (lrRNA-Seq) using Oxford Nanopore Technologies (ONT) or Pacific Biosciences (PacBio) platforms produces full-length transcript reads that enable isoform-level analysis, but the quality control (QC) steps differ substantially from short-read workflows. This article provides a practical pipeline for QC of lrRNA-Seq data, covering raw read assessment, adapter and poly(A) tail checking, read filtering, mapping QC, and isoform quantification readiness, with specific attention to direct RNA versus cDNA library differences.

The core problem researchers face is that standard short-read QC tools do not address transcriptomic features unique to long reads, such as poly(A) tail integrity, adapter contamination at both ends, splice junction accuracy, and the distinction between biological isoform variation and technical artifacts. A robust QC pipeline must integrate platform-specific metrics, library preparation knowledge, and downstream analysis requirements to ensure that isoform detection and quantification are biologically meaningful.

Key Takeaways

• Long-read RNA sequencing quality control requires platform-specific checks for read length distribution, adapter contamination at both ends, poly(A) tail integrity, and splice junction accuracy that differ from short-read workflows.
• The QC pipeline stages include raw read assessment with NanoPlot or pycoQC, adapter and poly(A) tail checking with Porechop, read filtering with Chopper or Filtlong, and mapping QC with minimap2 or STARlong.
• Direct RNA libraries require verifying poly(A) tail presence and length because the sequencing adapter ligates to the poly(A) tail, while cDNA libraries need checks for reverse transcription artifacts and PCR duplication.
• PacBio HiFi reads achieve higher accuracy through circular consensus sequencing, and a systematic evaluation found PacBio outperformed ONT in accuracy of novel transcript identification and allele-specific expression analysis.
• SQANTI3 classifies transcript models into structural categories for isoform quantification readiness, and researchers should document platform choice and version information because these factors influence interpretation of downstream results.

Scope and Reader Context

This pipeline is designed for biology students, researchers, laboratory professionals, and life-science practitioners who have generated or plan to generate lrRNA-Seq data. The workflow assumes familiarity with basic command-line operations and Unix-based computing environments. The Carpentries offers foundational lessons in shell scripting, data management, and programming that are useful prerequisites for implementing the commands described here [<a href="#ref-1">1</a>].

The pipeline covers the complete QC journey from raw sequencing output through to files ready for isoform quantification. It addresses both ONT and PacBio platforms, though the specific tools and parameters differ. Direct RNA sequencing, which sequences native RNA molecules without a cDNA conversion step, requires different QC considerations than cDNA-based libraries. The workflow presented here accommodates both library types with appropriate adjustments at each stage.

Core Principles of Long-Read RNA-Seq Quality Control

Why Long-Read QC Differs from Short-Read QC

Short-read sequencing generates fragments of 150 to 300 base pairs, and QC focuses primarily on base call accuracy, adapter contamination, and duplication rates. Long-read sequencing produces reads that can span entire transcripts, often exceeding 10 kilobases. This fundamental difference means that QC must assess features that short-read pipelines never encounter.

The primary QC dimensions for lrRNA-Seq include read length distribution, which directly reflects transcript integrity, base call quality scores along the read, which vary by position and platform, adapter contamination at both read ends, which affects mapping and quantification, poly(A) tail presence and length, which is critical for direct RNA libraries, and splice junction accuracy, which determines whether detected isoforms are real or artifacts.

Alternative splicing contributes substantially to biological heterogeneity between species, sexes, tissues, and cell types, and many diseases involve alterations in splicing patterns [<a href="#ref-2">2</a>]. Accurate isoform quantification therefore depends on QC procedures that distinguish genuine splice variants from sequencing errors or library preparation artifacts.

Platform-Specific Considerations

ONT and PacBio platforms generate long reads through fundamentally different technologies. ONT measures changes in electrical current as DNA or RNA passes through a protein nanopore, while PacBio uses zero-mode waveguides to detect fluorescent signals during synthesis. These technological differences produce distinct error profiles that QC pipelines must address.

ONT reads typically have lower per-base accuracy but can be generated in very high throughput, making them suitable for transcript discovery and quantification experiments. PacBio HiFi reads achieve higher accuracy through circular consensus sequencing, which improves novel transcript identification and allele-specific expression analysis [<a href="#ref-3">3</a>]. A systematic evaluation comparing both platforms on the same single-cell cDNA libraries found that PacBio outperformed ONT in accuracy of novel transcript identification and allele-specific gene and isoform expression [<a href="#ref-3">3</a>].

The choice of platform affects QC thresholds, error correction strategies, and the confidence assigned to detected isoforms. Researchers should document platform choice and version information in their analysis records, as these factors influence interpretation of downstream results.

Library Preparation Effects on QC Metrics

The library preparation method determines what QC checks are necessary and what artifacts may appear. cDNA libraries involve reverse transcription of RNA to cDNA, followed by PCR amplification in many protocols. Direct RNA libraries sequence native RNA molecules and require no amplification step.

For cDNA libraries, QC must check for reverse transcription artifacts, PCR duplication, and incomplete second-strand synthesis. For direct RNA libraries, QC must verify RNA integrity, check for truncated molecules, and assess poly(A) tail status since the sequencing adapter ligates to the poly(A) tail.

A study generating ONT lrRNA-Seq data from mouse brain cDNA libraries reported that over 85 million reads passed their quality control metrics, demonstrating that large-scale cDNA-based experiments can achieve acceptable QC outcomes [<a href="#ref-2">2</a>]. The same study used these QC-passed reads for differential gene expression, differential transcript expression, and differential transcript usage analyses across brain regions and sexes [<a href="#ref-2">2</a>].

At a Glance: QC Pipeline Overview

This table summarizes the main stages of the QC pipeline, the tools commonly used at each stage, the metrics that inform decisions, and the specific decision points where researchers must evaluate whether to proceed, filter, or re-sequence.

Practical Workflow: Step-by-Step QC Implementation

Step 1: Raw Data Assessment

The first stage of any lrRNA-Seq QC pipeline is assessing the raw sequencing output before any filtering or trimming. This provides a baseline understanding of data volume, read length distribution, and overall quality.

NanoPlot generates comprehensive summary statistics and visualizations for ONT data, including read length distributions, quality score distributions, and cumulative yield plots. pycoQC provides similar functionality with interactive plots that can be generated from the sequencing summary file produced by ONT instruments. For PacBio data, the SMRT Link software suite includes QC reporting tools that summarize read length, quality, and yield.

Key metrics to record at this stage include total read count, total bases, read length N50, mean read quality, and the distribution of read lengths. The N50 value is particularly informative because it indicates the read length at which half of all bases are contained in reads of that length or longer. For transcriptomic analysis, higher N50 values generally indicate better preservation of full-length transcripts.

The number and distribution of reads, as well as the number and distribution of unique junction chains, are informative of raw data quality in replicated lrRNA-Seq experiments [<a href="#ref-4">4</a>]. Junction chains refer to the specific combinations of splice junctions observed in individual reads, and their diversity reflects the complexity of isoform expression in the sample.

Step 2: Adapter Contamination Checking

Adapter sequences are added during library preparation and must be removed before mapping and quantification. In lrRNA-Seq, adapters can appear at both the 5' and 3' ends of reads, and incomplete adapter removal can cause reads to fail mapping or map incorrectly.

Porechop is a common tool for detecting and removing adapter sequences from ONT reads. It identifies adapter sequences at read ends and trims them accordingly. For PacBio data, adapter removal is typically performed during the circular consensus sequencing step, but residual adapters can still appear and should be checked.

Custom scripts can complement automated adapter trimming by searching for known adapter sequences in read ends and reporting the frequency and position of adapter contamination. This is particularly useful when using non-standard adapter sequences or when adapting protocols for specific applications.

For direct RNA libraries, the poly(A) tail serves as the attachment point for the sequencing adapter. The QC pipeline must distinguish between the poly(A) tail, which is a biological feature of mRNA, and adapter sequences, which are technical artifacts. Trimming the poly(A) tail is generally recommended before mapping, but the tail length should be recorded because it can provide information about RNA integrity and polyadenylation status.

Step 3: Poly(A) Tail Assessment

Poly(A) tail assessment is a critical QC step for direct RNA sequencing and is also relevant for cDNA libraries that retain poly(A) information. The poly(A) tail length varies across transcripts and can influence sequencing efficiency and downstream analysis.

For direct RNA libraries, the presence and length of poly(A) tails can be assessed using tools such as Nanopolish or custom scripts that identify homopolymer stretches at read ends. The distribution of poly(A) tail lengths across reads provides information about RNA quality and the efficiency of the library preparation.

For cDNA libraries, poly(A) tails may be present if the reverse transcription primer annealed within the poly(A) tail instead of at its 5' boundary. These internal priming events can produce reads that appear to have poly(A) tails but actually represent truncated transcripts. QC should flag reads with unusually long poly(A) tails for further investigation.

The decision to trim poly(A) tails before mapping depends on the downstream analysis. Some isoform quantification tools handle poly(A) tails internally, while others require them to be removed. Researchers should consult the documentation for their chosen analysis tools to determine the appropriate preprocessing steps.

Step 4: Read Filtering

Read filtering removes reads that are too short, have low quality, or are likely to be artifacts. The specific thresholds depend on the experimental goals and the platform used.

For transcriptomic analysis, a minimum read length threshold is typically applied to remove short fragments that cannot provide meaningful isoform information. The threshold should be informed by the expected transcript length distribution in the sample. For example, a study focused on full-length mRNA isoforms might set a minimum length of 500 bases, while a study including small RNAs would use a lower threshold.

Quality score thresholds remove reads with high error rates that could introduce spurious splice junctions or base substitutions. The appropriate threshold depends on the platform and the downstream analysis. Higher accuracy platforms such as PacBio HiFi can tolerate more aggressive quality filtering, while ONT data may require more lenient thresholds to retain sufficient read depth.

Chopper and NanoFilt are commonly used for filtering ONT reads by length and quality. Filtlong provides additional filtering options based on read quality and length. For PacBio data, filtering is often integrated into the SMRT Link analysis pipeline.

The filtering stage should produce a clear record of how many reads were removed and why. This information is essential for reproducibility and for interpreting downstream results.

Step 5: Mapping and Alignment QC

After filtering, reads are mapped to a reference genome or transcriptome. The mapping step is where many QC issues become apparent, including adapter contamination that survived trimming, chimeric reads, and mapping artifacts.

minimap2 is a widely used aligner for long reads and supports both genome and transcriptome mapping. STARlong is an alternative that extends the STAR aligner for long reads. The choice of aligner and mapping parameters affects mapping rates and the accuracy of splice junction detection.

Mapping QC metrics include the overall mapping rate, the distribution of reads across genomic features, and the number of reads mapping to unexpected locations. Low mapping rates may indicate adapter contamination, sample contamination, or reference genome issues.

SQANTI3 provides a framework for classifying transcript models into structural categories based on their relationship to annotated transcripts. SQANTI-reads extends this framework to read-level QC, leveraging SQANTI3 categories to assess the quality of replicated lrRNA-Seq experiments [<a href="#ref-4">4</a>]. The number and distribution of reads and unique junction chains across SQANTI3 structural categories are informative of raw data quality [<a href="#ref-4">4</a>].

Multisample visualizations of QC metrics presented by experimental design factors can identify outliers in replicated experiments [<a href="#ref-4">4</a>]. This is particularly useful for detecting batch effects or sample processing issues that might not be apparent from single-sample QC.

Step 6: Splice Junction and Isoform Structure QC

Splice junction accuracy is a critical QC dimension for lrRNA-Seq because the primary advantage of long reads is their ability to capture complete isoform structures. Errors in splice junction detection can lead to false isoform calls and incorrect quantification.

SQANTI-reads introduces metrics for identifying potentially under-annotated genes and putative novel transcripts, as well as for quantifying variation in junction donors and acceptors [<a href="#ref-4">4</a>]. These metrics help distinguish genuine novel isoforms from artifacts caused by sequencing errors or mapping errors.

The tool effectively reveals the impact of read coverage on data quality and readily identifies strong and weak splicing sites [<a href="#ref-4">4</a>]. Weak splicing sites may indicate sequencing errors or may represent genuine alternative splicing events that are difficult to detect with lower coverage.

For replicated experiments, the consistency of junction chain distributions across replicates provides information about technical variability. High variability between replicates may indicate library preparation issues, sequencing depth differences, or biological variation.

Step 7: Isoform Quantification Readiness Assessment

The final QC stage determines whether the data is ready for isoform quantification. This assessment integrates information from all previous stages and evaluates whether the data meets the requirements of the chosen quantification tool.

Isoform quantification tools such as FLAIR, IsoQuant, and Salmon (in long-read mode) have specific input requirements and assumptions about data quality. Researchers should verify that their QC-passed data meets these requirements before proceeding.

Key readiness indicators include sufficient read depth for reliable isoform detection, acceptable mapping rates, reasonable junction chain diversity, and consistency across replicates. The specific thresholds depend on the experimental design and the biological questions being addressed.

For experiments focused on differential transcript usage, the QC pipeline should ensure that read coverage is sufficient across all conditions being compared. A study of mouse brain regions found that the cerebellum had the most differences compared to the other three regions examined, highlighting the importance of adequate coverage for detecting region-specific isoform differences [<a href="#ref-2">2</a>].

Tools and Their Roles in the QC Pipeline

NanoPlot and pycoQC for ONT Data

NanoPlot generates static plots and summary statistics from ONT sequencing data, including read length distributions, quality score distributions, and yield plots. It accepts multiple input formats, including FASTQ, FASTA, and BAM files, and can generate both summary tables and publication-ready figures.

pycoQC provides interactive plots generated from the sequencing summary file produced by ONT instruments. It offers real-time QC during sequencing runs and can generate reports for completed runs. The interactive nature of pycoQC plots makes them useful for exploring data distributions and identifying anomalies.

Both tools are available through the Bioconductor ecosystem, which provides comprehensive documentation and reproducible analysis workflows [<a href="#ref-5">5</a>]. Bioconductor packages follow standardized installation and usage patterns, making them accessible to researchers with basic R programming skills.

Porechop and cutadapt for Adapter Trimming

Porechop identifies and removes adapter sequences from ONT reads. It can detect adapters at both read ends and can handle the specific adapter configurations used in different library preparation kits. Porechop also provides information about the frequency of adapter contamination, which can indicate library preparation issues.

cutadapt is a general-purpose adapter trimming tool that works with both ONT and PacBio data. It supports a wide range of adapter configurations and can be used in custom pipelines where more control over trimming parameters is needed.

The choice between Porechop and cutadapt depends on the specific adapter sequences used and the level of control required. Porechop is optimized for ONT adapters and provides automatic detection, while cutadapt offers more flexibility for custom adapter sequences.

SQANTI3 and SQANTI-reads for Transcript Structure QC

SQANTI3 classifies transcript models into structural categories based on their relationship to annotated transcripts. These categories include full splice match, incomplete splice match, novel in catalog, novel not in catalog, and others. The distribution of transcripts across these categories provides information about the quality of transcript reconstruction.

SQANTI-reads extends SQANTI3 to read-level QC, providing a framework for assessing the quality of replicated lrRNA-Seq experiments [<a href="#ref-4">4</a>]. It introduces metrics for identifying potentially under-annotated genes and putative novel transcripts, as well as for quantifying variation in junction donors and acceptors [<a href="#ref-4">4</a>].

The tool provides multisample visualizations of QC metrics presented by experimental design factors to identify outliers [<a href="#ref-4">4</a>]. This is particularly useful for large experiments with multiple conditions and replicates.

Custom Scripts for Specific QC Tasks

Custom scripts are often necessary for QC tasks that are not covered by existing tools. These may include checking for specific adapter sequences, assessing poly(A) tail length distributions, or generating custom visualizations.

The Galaxy Training Network provides accessible workflow training and analysis tutorials that can help researchers develop custom QC scripts within a reproducible framework [<a href="#ref-6">6</a>]. Galaxy workflows can be shared and reused, making them useful for standardizing QC procedures across projects.

nf-core documentation describes community pipeline standards and usage, providing a framework for developing reproducible QC pipelines [<a href="#ref-7">7</a>]. nf-core pipelines follow standardized configuration and usage patterns, making them accessible to researchers with basic command-line experience.

Direct RNA versus cDNA Library QC Considerations

Direct RNA Sequencing QC

Direct RNA sequencing sequences native RNA molecules without reverse transcription or amplification. This approach preserves RNA modifications and provides information about poly(A) tail length, but it has lower throughput and higher error rates than cDNA-based approaches.

QC for direct RNA data must assess RNA integrity, which is reflected in read length distribution. Degraded RNA produces shorter reads and reduces the proportion of full-length transcripts. The poly(A) tail is a critical QC feature for direct RNA data because the sequencing adapter ligates to the tail, and tail length affects sequencing efficiency.

Direct RNA libraries do not require adapter trimming at the 5' end because the adapter is added during library preparation. However, the 3' end may contain adapter sequences that need to be removed. The poly(A) tail should be trimmed before mapping unless the downstream analysis specifically requires tail information.

cDNA Library QC

cDNA libraries involve reverse transcription of RNA to cDNA, followed by PCR amplification in most protocols. The QC considerations differ from direct RNA sequencing because the library preparation introduces additional artifacts.

Reverse transcription can produce truncated cDNAs if the enzyme falls off before reaching the 5' end of the transcript. These truncated molecules appear as shorter reads and can bias isoform quantification toward the 3' ends of transcripts. PCR amplification can introduce duplication artifacts and chimeric molecules.

QC for cDNA libraries should assess the proportion of reads that span full-length transcripts, the duplication rate, and the presence of chimeric reads. Tools that classify reads by their relationship to annotated transcripts, such as SQANTI3, are particularly useful for identifying these artifacts.

A study using ONT lrRNA-Seq with cDNA libraries from mouse brain generated over 85 million reads that passed quality control metrics, demonstrating that cDNA-based approaches can achieve sufficient depth for differential expression and isoform usage analyses [<a href="#ref-2">2</a>].

Records and Measurements for QC Documentation

Essential Records for Reproducibility

Reproducibility requires detailed records of all QC steps, parameters, and decisions. The following records should be maintained for each sequencing run and analysis:

Sequencing run metadata including platform, flow cell version, chemistry version, and run date. Library preparation records including library type, input RNA amount, reverse transcription protocol, PCR cycle number, and adapter sequences. QC tool versions and parameters for every tool used in the pipeline. Filtering decisions including the number of reads removed at each stage and the reasons for removal. Mapping statistics including mapping rate, read length distribution after mapping, and junction chain counts.

The Carpentries lessons emphasize the importance of reproducible data analysis practices, including version control, documentation, and automation [<a href="#ref-1">1</a>]. These practices are essential for QC pipelines that may be rerun with different parameters or applied to new datasets.

Metrics to Track Across the Pipeline

The following metrics should be tracked at each stage of the QC pipeline:

Read count and total bases at each stage. Read length N50 and mean read length. Mean quality score and quality score distribution. Adapter contamination rate before and after trimming. Poly(A) tail length distribution for direct RNA data. Mapping rate and reasons for unmapped reads. Junction chain count and distribution across structural categories. Replicate consistency metrics for multi-sample experiments.

The number and distribution of reads and unique junction chains in SQANTI3 structural categories are informative of raw data quality [<a href="#ref-4">4</a>]. Tracking these metrics across the pipeline helps identify where data quality issues originate.

Outlier Detection in Multi-Sample Experiments

For experiments with multiple samples or replicates, QC should include outlier detection to identify samples that differ substantially from the group. SQANTI-reads provides multisample visualizations of QC metrics presented by experimental design factors to identify outliers [<a href="#ref-4">4</a>].

Outlier samples may indicate library preparation issues, sequencing problems, or biological variation. The decision to exclude an outlier sample should be documented and justified based on the QC metrics and the experimental context.

In replicated experiments, the consistency of QC metrics across replicates provides information about technical variability. High variability may indicate the need for additional quality control steps or may reflect genuine biological differences between samples.

Common Failure Patterns in Long-Read RNA-Seq QC

Adapter Contamination Persisting After Trimming

Adapter contamination can persist after trimming if the adapter sequences are not correctly identified or if the trimming parameters are too conservative. This is particularly common with custom adapter sequences or when using library preparation kits with non-standard adapter configurations.

Symptoms of persistent adapter contamination include reduced mapping rates, reads that map to unexpected locations, and the appearance of chimeric transcripts. The QC pipeline should include a post-trimming adapter check to verify that contamination has been removed.

Poly(A) Tail Artifacts in cDNA Libraries

Poly(A) tails in cDNA libraries can indicate internal priming during reverse transcription. This occurs when the oligo(dT) primer anneals to an internal poly(A) stretch within the transcript instead of to the poly(A) tail at the 3' end.

Internal priming produces reads that appear to have poly(A) tails but actually represent truncated transcripts. These reads can bias isoform quantification and should be flagged during QC. Custom scripts that identify reads with poly(A) tails at internal positions can help identify this artifact.

Chimeric Reads and Mapping Artifacts

Chimeric reads are reads that contain sequences from two different transcripts or genomic locations. They can arise from PCR artifacts during library preparation or from errors during sequencing.

Chimeric reads can produce false isoform calls and incorrect quantification. The QC pipeline should include checks for chimeric reads, such as examining reads that map to multiple genomic locations or that have unexpected junction patterns.

Low Mapping Rates

Low mapping rates can indicate several problems, including adapter contamination, sample contamination, reference genome issues, or sequencing errors. The QC pipeline should investigate the causes of low mapping rates before proceeding with downstream analysis.

For transcriptomic data, mapping to the genome versus the transcriptome can produce different mapping rates. Genome mapping is more comprehensive but may be more sensitive to sequencing errors, while transcriptome mapping is faster but may miss novel isoforms.

Replicate Inconsistency

Inconsistent QC metrics across replicates can indicate technical issues or biological variation. The QC pipeline should include statistical tests for replicate consistency and should flag samples that deviate substantially from the group.

The impact of read coverage on data quality is readily revealed by tools such as SQANTI-reads, which can identify samples with insufficient coverage for reliable isoform detection [<a href="#ref-4">4</a>]. Low coverage samples may need to be re-sequenced or excluded from downstream analysis.

Limitations of Long-Read RNA-Seq QC

Error Rate and Isoform Detection Accuracy

Long-read sequencing has higher error rates than short-read sequencing, which can affect isoform detection accuracy. The impact of errors depends on the platform, the error profile, and the downstream analysis.

PacBio HiFi sequencing achieves higher accuracy than ONT sequencing, and a systematic evaluation found that PacBio outperformed ONT in accuracy of novel transcript identification and allele-specific gene and isoform expression [<a href="#ref-3">3</a>]. Researchers using ONT data should be aware that higher error rates may reduce confidence in novel isoform calls.

Coverage Requirements for Reliable Quantification

Reliable isoform quantification requires sufficient read coverage for each transcript. Low coverage can lead to false negative isoform calls and inaccurate quantification.

The coverage required depends on the complexity of the transcriptome, the abundance of specific isoforms, and the goals of the analysis. SQANTI-reads can reveal the impact of read coverage on data quality, helping researchers determine whether additional sequencing is needed [<a href="#ref-4">4</a>].

Annotation Dependence

Many QC tools rely on reference annotations to classify transcripts and assess quality. Incomplete or inaccurate annotations can lead to misclassification of novel isoforms and incorrect QC conclusions.

The identification of potentially under-annotated genes and putative novel transcripts is a specific focus of SQANTI-reads, which provides metrics for these features [<a href="#ref-4">4</a>]. Researchers working with poorly annotated genomes should interpret QC results with caution.

Platform-Specific Limitations

Each sequencing platform has specific limitations that affect QC. ONT data has higher error rates and may require more aggressive filtering, while PacBio data has higher accuracy but lower throughput.

The choice of platform affects the types of analyses that can be reliably performed. A systematic evaluation found that while both ONT and PacBio can be used for single-cell RNA sequencing, PacBio outperformed ONT in accuracy of novel transcript identification and allele-specific gene and isoform expression [<a href="#ref-3">3</a>].

Safety and Regulatory Context for RNA-Seq Data

Data Management and Privacy Considerations

RNA-seq data may contain sensitive information, particularly if derived from human samples. Researchers should follow institutional guidelines for data storage, sharing, and privacy protection.

The NCBI provides data resources and search systems for sequence data, including databases for raw sequencing reads and processed expression data [<a href="#ref-8">8</a>]. Researchers depositing data in public repositories should ensure that data is properly de-identified and that appropriate consent has been obtained.

Reproducibility Standards

Reproducibility is a core requirement for published research. The QC pipeline should be documented in sufficient detail that other researchers can reproduce the analysis.

GTN's training modules guide users through cloud-based workflow execution environments that support provenance tracking and result reproducibility.

Professional Escalation Criteria

Researchers should escalate QC issues to appropriate professionals when they cannot resolve problems independently. The following situations warrant escalation:

Persistent low mapping rates that cannot be explained by known issues. Unexpected patterns in QC metrics that suggest systematic problems. Discrepancies between replicates that cannot be resolved by filtering. Evidence of sample contamination or mix-ups. Issues with reference genome or annotation quality that affect QC interpretation.

Institutional bioinformatics cores, sequencing facility staff, and collaboration with experienced bioinformaticians can provide support for resolving complex QC issues.

Practical Implementation Steps

Setting Up the QC Environment

The QC pipeline requires a computing environment with the necessary tools installed. Bioconductor provides official package, workflow, installation, and reproducible genomic-analysis documentation [<a href="#ref-5">5</a>]. The Carpentries offers foundational lessons in shell, Git, and programming that are useful for setting up and managing analysis environments [<a href="#ref-1">1</a>].

Maintaining complete pipeline execution logs - including software versions, input checksums, and configuration parameters - is critical for diagnosing discrepancies between analysis runs.

Running the Pipeline

The QC pipeline should be run in a consistent order, with clear documentation of parameters and decisions at each stage. The pipeline can be automated using workflow management systems such as nf-core, which provides community pipeline standards and usage documentation [<a href="#ref-7">7</a>].

For multi-sample experiments, the pipeline should be run consistently across all samples, and QC metrics should be compared across samples to identify outliers.

Interpreting QC Results

QC results should be interpreted in the context of the experimental design and the biological questions being addressed. A QC metric that is acceptable for one experiment may be unacceptable for another.

The interpretation of QC results should be documented, including the rationale for any filtering decisions. This documentation is essential for reproducibility and for defending analysis decisions in publications.

Documenting QC Decisions

All QC decisions should be documented, including the parameters used, the number of reads removed at each stage, and the rationale for filtering thresholds. This documentation should be included in the analysis records and referenced in publications.

The documentation should include the versions of all tools used, the dates of analysis, and any deviations from standard protocols. This information is essential for reproducing the analysis and for troubleshooting issues.

Frequently Asked Questions

What is the minimum read length for long-read RNA-seq analysis?

The minimum read length depends on the experimental goals and the expected transcript length distribution. For full-length isoform analysis, reads should be long enough to span complete transcripts. A common approach is to set a minimum length threshold based on the expected transcript lengths in the sample, often in the range of 500 to 1000 bases. Shorter reads may still be useful for gene-level analysis but provide limited isoform information.

How do I check for adapter contamination in long reads?

Adapter contamination can be checked using tools such as Porechop for ONT data or cutadapt for general adapter trimming. These tools identify adapter sequences at read ends and can report the frequency of contamination. Custom scripts can also search for known adapter sequences and report their positions in reads. Post-trimming checks should verify that adapter contamination has been removed.

Should I trim poly(A) tails before mapping?

Poly(A) tail trimming is generally recommended before mapping, particularly for direct RNA data. The poly(A) tail can interfere with mapping and can cause reads to map incorrectly. However, some downstream analysis tools handle poly(A) tails internally, so the decision depends on the specific tools being used. The poly(A) tail length should be recorded before trimming because it provides information about RNA integrity.

What mapping rate should I expect for long-read RNA-seq data?

Mapping rates vary depending on the platform, library type, reference genome quality, and filtering parameters. High-quality data typically achieves mapping rates above 80 percent, but lower rates may be acceptable for certain experiments. Low mapping rates should be investigated to determine the cause, which may include adapter contamination, sample contamination, or reference genome issues.

How do I identify novel isoforms in long-read RNA-seq data?

Novel isoform identification requires comparing detected transcripts to reference annotations. Tools such as SQANTI3 classify transcripts into structural categories, including novel isoforms. SQANTI-reads provides metrics for identifying potentially under-annotated genes and putative novel transcripts [<a href="#ref-4">4</a>]. Novel isoform calls should be validated with additional evidence, particularly for platforms with higher error rates.

What is the difference between direct RNA and cDNA library QC?

Direct RNA sequencing sequences native RNA molecules and requires QC checks for RNA integrity and poly(A) tail status. cDNA libraries involve reverse transcription and PCR amplification, which introduce different artifacts such as truncated cDNAs and chimeric molecules. The QC pipeline must be adjusted based on the library type to address the specific artifacts that can occur.

How many reads do I need for reliable isoform quantification?

The number of reads needed depends on the transcriptome complexity, the abundance of specific isoforms, and the goals of the analysis. SQANTI-reads can reveal the impact of read coverage on data quality, helping researchers determine whether additional sequencing is needed [<a href="#ref-4">4</a>]. For replicated experiments, consistency across replicates provides information about whether coverage is sufficient.

How do I handle outlier samples in multi-sample experiments?

Outlier samples should be identified using multisample QC visualizations that present metrics by experimental design factors [<a href="#ref-4">4</a>]. Once identified, outliers should be investigated to determine whether they represent technical issues or biological variation. The decision to exclude an outlier should be documented and justified based on the QC metrics and experimental context.

Related Bioinformatics Guides

• RNA-Seq Quality Control: Essential Checks and Tools
• RNA-Seq Data Analysis Workflow: From Raw Reads to Insights
• Long-Read Sequencing for Isoform Quantification: Challenges and Solutions
• RNA Sequencing Data Analysis: From Raw Reads to Differential Expression
• Single-Cell RNA Sequencing Quality Control: A Practical Guide to Filtering and Metrics

References and Further Reading

This article is educational and does not replace validated analysis plans, institutional policy, clinical interpretation, or specialist review.