# A Decision Guide to Choosing Between FASTQ and BAM for Long-Read Data Storage and Analysis Researchers working with long-read sequencing data from platforms such as Oxford Nanopore and Pacific Biosciences face a practical decision at the point of data generation: whether to store and analyze sequencing output as raw FASTQ files or as aligned BAM files. This decision affects storage costs, compute time, analysis flexibility, and the ability to revisit data as new tools and reference genomes become available. The direct answer is that most projects benefit from keeping both formats, with FASTQ as the archival master copy and BAM as the working analysis format, but the optimal balance depends on your specific research questions, institutional storage limits, and downstream analysis plans. This article provides a structured framework for making that decision based on your project goals, available infrastructure, and long-term data management needs. ## Understanding the Two Core Formats FASTQ and BAM serve fundamentally different purposes in the bioinformatics pipeline. FASTQ is the primary output format from sequencing instruments and contains the raw nucleotide sequence reads along with per-base quality scores. BAM is a compressed binary alignment format that contains the same sequence reads plus their alignment positions against a reference genome. Understanding what each format preserves and what it discards is essential for making an informed storage decision. ### FASTQ as the Raw Data Record FASTQ files store four lines per read: a sequence identifier, the nucleotide sequence, a separator line, and quality scores encoded as ASCII characters. For long-read platforms, FASTQ files are typically large because reads can span tens of kilobases. The format is platform-agnostic and does not depend on any particular reference genome. This reference-independence is a critical property for archival storage because it means the data can be realigned to improved reference assemblies or different reference species without loss of information. The NCBI Sequence Read Archive accepts FASTQ files as a primary submission format for raw sequencing data, and this repository serves as the standard destination for public data deposition. When you submit to NCBI, you are preserving the raw read data in a form that other researchers can download and analyze with their own pipelines. The [NCBI data resources](https://www.ncbi.nlm.nih.gov/) provide documentation on accepted submission formats and database organization for sequence data. FASTQ files are text-based and can be compressed with standard tools such as gzip. The compression ratio for long-read FASTQ data is generally modest because sequence data is already fairly compact, but quality score lines compress well. A typical Nanopore run producing 10 to 20 gigabases of sequence data will generate FASTQ files in the range of 10 to 30 gigabytes before compression, depending on read length distribution and quality score encoding. ### BAM as the Analysis-Ready Format BAM files contain the same read sequences and quality scores as FASTQ but add alignment information: the reference sequence name, start position, mapping quality, and a compact representation of the alignment including insertions, deletions, and soft-clipped bases. The BAM format is binary and compressed by default, making it substantially smaller than the equivalent FASTQ file for aligned reads. The compression works because reads that align to similar genomic regions share sequence content, allowing the compression algorithm to exploit redundancy. The alignment information in BAM files enables immediate access to genomic coordinates, which is required for most downstream analyses including variant calling, structural variant detection, and transcript quantification. Tools in the [Bioconductor project](https://bioconductor.org/) provide extensive functionality for reading, manipulating, and analyzing BAM files within the R statistical environment, including packages for read counting, coverage visualization, and variant analysis. BAM files are reference-dependent. If you align reads to one reference genome and later want to use a newer reference, you must either realign from the original FASTQ files or use a tool that can lift coordinates between assemblies. Realignment from FASTQ is the more reliable approach because it avoids the coordinate conversion errors that can occur when lifting between references. ## At a Glance: FASTQ versus BAM for Long-Read Data | Decision Factor | FASTQ | BAM | |-----------------|-------|-----| | Primary content | Raw reads with quality scores | Aligned reads with genomic coordinates | | Reference genome required | No | Yes | | File size for long reads | Larger, especially before compression | Smaller due to binary compression and alignment redundancy | | Analysis flexibility | Can be realigned to any reference or used with reference-free tools | Restricted to the reference used for alignment | | Archival suitability | High, preserves all information | Moderate, alignment may become outdated | | Tool compatibility | Universal, accepted by all aligners and assemblers | Requires alignment-aware tools, most variant callers expect BAM | | Storage cost per project | Higher | Lower | | Compute cost to generate | None, produced by sequencer | Requires alignment step | | Best use case | Long-term archival and data sharing | Active analysis and intermediate results | ## Core Principles for Format Selection The decision between FASTQ and BAM storage is governed by several principles that apply across different project types. These principles help researchers balance the competing demands of data preservation, analysis efficiency, and storage economy. ### Principle One: Preserve the Most Information-Dense Format The first principle is that raw FASTQ data is the most information-dense format available for sequencing data. It contains every base call and quality score produced by the instrument, with no reference-dependent interpretation. Any analysis that can be performed on BAM files can also be performed on FASTQ files after an alignment step, but the reverse is not true. Once you discard FASTQ files, you cannot recover information that was lost during alignment, such as reads that failed to map or reads that map to multiple locations. For projects where data may be reanalyzed in the future with improved tools, the FASTQ files are the insurance policy. The [Galaxy Training Network](https://training.galaxyproject.org/) emphasizes reproducible analysis workflows, and reproducibility starts with preserving the raw data in a format that future pipelines can consume. A BAM file created with an older aligner may not be compatible with newer variant callers that expect specific alignment tags or quality metrics. ### Principle Two: Match Storage Format to Analysis Stage The second principle is that different analysis stages have different format requirements. During active analysis, BAM files are the practical working format because they enable fast access to genomic regions and are required by most variant calling and quantification tools. During project archival and data sharing, FASTQ files are the appropriate format because they preserve maximum flexibility for future analyses. A practical workflow is to maintain FASTQ files as the master copy, generate BAM files during analysis, and delete intermediate BAM files when they are no longer needed. The BAM files can always be regenerated from FASTQ if the alignment parameters are documented. This approach minimizes storage costs while preserving the ability to reproduce and extend analyses. ### Principle Three: Consider the Cost of Regeneration The third principle is that storage decisions should account for the cost of regenerating derived data. If you delete BAM files but keep FASTQ files, you can regenerate the BAM files with an alignment step. The compute cost of alignment for long-read data is substantial, often requiring several hours per sample on a multi-core server. If you anticipate needing the BAM files again, it may be more cost-effective to store them than to regenerate them. Conversely, if you delete FASTQ files and keep only BAM files, you cannot regenerate the raw reads. Any analysis that requires unaligned reads, such as de novo assembly or reference-free transcript quantification, becomes impossible. The asymmetry in regeneration cost strongly favors keeping FASTQ files as the archival format. ### Principle Four: Benchmark Before Committing to a Workflow The fourth principle is that tool and format choices should be validated with systematic benchmarking before committing to a large-scale analysis. A study on SARS-CoV-2 subgenomic RNA detection demonstrated that common bioinformatics tools showed substantial performance variability depending on the aligner chosen and the sequencing strategy used, with some tools struggling on certain data types and being sensitive to mutations depending on the aligner. The authors emphasized that without systematic benchmarking, researchers risk drawing inaccurate conclusions from suboptimal workflows. This finding applies directly to format decisions: the choice between FASTQ and BAM storage should be informed by testing your specific downstream tools on representative data to understand which format they require and how alignment choices affect results. The [benchmarking study](https://doi.org/10.3389/fbinf.2026.1803237) highlights the value of context-aware tool selection and standardized benchmarking practices for ensuring reproducibility and reliability in bioinformatics analysis. ## Practical Workflow for Long-Read Data Management Implementing a format strategy requires a concrete workflow that integrates with your existing analysis pipeline. The following steps provide a framework for managing long-read data from instrument output through archival storage. ### Step One: Define Your Data Tiers Before data generation begins, define three data tiers with different retention policies. The first tier is the raw instrument output, which includes FASTQ files and any platform-specific files such as basecalling intermediates. The second tier is the analysis-ready data, which includes BAM files and derived analysis results. The third tier is the archival data, which is the minimal set of files needed to reproduce all published results. For most projects, the archival tier should include the FASTQ files, the reference genome version, the aligner and parameters used, and the analysis scripts. The [nf-core documentation](https://nf-co.re/docs) describes how community-developed pipelines handle data management and reproducibility, and these practices can inform your own data management plan. Community pipelines typically accept FASTQ files as input and generate BAM files as intermediate output, reflecting the standard practice of treating FASTQ as the primary input format. ### Step Two: Establish Naming and Directory Conventions Consistent file naming and directory structure are essential for managing large numbers of sequencing files. Use a naming convention that includes the sample identifier, the sequencing platform, the read type, and the file format. For example, a file named `sample01_nanopore_rna.fastq.gz` clearly indicates the sample, platform, and data type. Organize directories by project, then by sample, then by data tier. Keep FASTQ files in a separate directory from BAM files to prevent accidental overwriting and to simplify backup procedures. Document the directory structure in a README file that accompanies the data, following the principles of [The Carpentries lessons](https://carpentries.org/lessons) on reproducible research practices. ### Step Three: Implement Quality Control at Each Stage Quality control should be performed at multiple points in the workflow. When FASTQ files are generated, check read length distributions, quality score distributions, and total yield against expected values for the platform and library type. After alignment, check mapping rates, coverage uniformity, and insert size distributions for paired-end data. The [EMBL-EBI training resources](https://www.ebi.ac.uk/training) provide structured learning pathways for quality assessment and data handling in bioinformatics. These resources emphasize the importance of understanding data quality before proceeding with downstream analysis, and the same principle applies to data management decisions. Low-quality runs may warrant different storage decisions than high-quality runs, such as discarding failed samples earlier in the workflow. ### Step Four: Document Alignment Parameters When you generate BAM files from FASTQ, document the exact aligner version, reference genome version, and all non-default parameters. This documentation is essential for reproducing the alignment and for understanding why BAM files may differ between analysis runs. Store this information in a machine-readable format such as a YAML or JSON file that accompanies the BAM files. The [Bioconductor project](https://bioconductor.org/) provides tools for managing and documenting genomic analyses within R, and many of these tools can generate session information that records package versions and analysis parameters. This level of documentation supports the reproducibility standards that are increasingly expected in genomics research. ### Step Five: Plan for Data Submission If your project will result in public data deposition, plan for submission to NCBI early in the project. The [NCBI data resources](https://www.ncbi.nlm.nih.gov/) accept raw sequencing data in FASTQ format for the Sequence Read Archive, and submission requirements include metadata about the sequencing platform, library preparation, and sample characteristics. Preparing submission files as part of the data management workflow reduces the burden at the end of the project. For projects that generate both raw and processed data, consider what derived data should be submitted alongside the FASTQ files. Some journals and repositories require processed data such as count matrices or variant call files, and these can be generated from BAM files during the analysis phase. ## Options and Tradeoffs in Storage Strategies Different storage strategies have different tradeoffs in terms of cost, flexibility, and risk. The following options represent common approaches used in research laboratories and core facilities. ### Option One: Store Everything The simplest strategy is to store all FASTQ and BAM files for the duration of the project. This approach maximizes flexibility and minimizes the risk of losing data that may be needed for reanalysis. The cost is the highest storage requirement, which can be substantial for large long-read projects. For projects with generous storage allocations or institutional data storage services, this approach is the safest. It eliminates the need to make decisions about which files to retain and simplifies the process of revisiting analyses with new tools. The main risk is that storage costs grow linearly with data generation, and projects that generate data continuously may exhaust available storage. ### Option Two: Store FASTQ Only The second strategy is to store only FASTQ files and regenerate BAM files as needed. This approach minimizes storage costs while preserving the ability to perform any analysis. The tradeoff is the compute cost of realignment, which can be significant for large datasets. This strategy works well for projects where analyses are performed infrequently or where the same FASTQ data is used for multiple different analyses with different alignment parameters. It also works well for projects that are still in the exploratory phase, where the optimal analysis approach has not yet been determined. ### Option Three: Store BAM Only The third strategy is to store only BAM files and discard FASTQ files after alignment. This approach minimizes storage costs but sacrifices the ability to realign to new references or use reference-free analysis tools. It is the riskiest strategy because it permanently loses the raw data. This strategy is only appropriate for projects where the reference genome is stable, the analysis approach is well-established, and there is no expectation of future reanalysis with different tools. It is not recommended for most research projects because the field of long-read analysis is evolving rapidly, and new tools frequently require different alignment approaches. ### Option Four: Tiered Storage with Compression The fourth strategy is to use tiered storage with different compression levels for different data types. FASTQ files can be compressed with standard gzip compression, which typically achieves a 3 to 4 fold reduction in file size. BAM files are already compressed, but additional compression can be applied for archival storage. Some research groups use specialized compression tools that achieve better compression ratios for sequencing data by exploiting the structure of the data. These tools can reduce storage requirements by an additional 20 to 50 percent compared to standard compression, but they require additional compute time and may not be compatible with all downstream tools. ## Observations and Measurements for Format Decisions Making informed format decisions requires measuring the actual storage and compute costs for your specific data. The following measurements provide the data needed to evaluate different storage strategies. ### Measuring File Sizes The first measurement is the file size of FASTQ and BAM files for representative samples. For long-read data, the ratio of BAM to FASTQ file size depends on the alignment rate and the read length distribution. Highly repetitive genomes or samples with substantial contamination will have lower alignment rates, resulting in BAM files that are closer in size to FASTQ files. Record the file sizes for each sample in a spreadsheet or database, along with the sequencing platform, library type, and alignment parameters. This data allows you to estimate storage requirements for future projects and to identify samples that deviate from expected size distributions. ### Measuring Alignment Compute Time The second measurement is the compute time required for alignment. This depends on the aligner used, the number of CPU cores available, the read length, and the genome size. Record the wall-clock time and CPU time for each alignment job, along with the aligner version and parameters. This data is essential for estimating the cost of regenerating BAM files from FASTQ. If alignment takes 10 hours per sample on your compute infrastructure, the cost of deleting BAM files and regenerating them later is substantial. If alignment takes 30 minutes per sample, the cost is minimal. ### Measuring Storage Costs The third measurement is the actual storage cost for your institution. This may be expressed as a dollar amount per terabyte per month, or it may be an allocation that is shared across the laboratory. Understanding the marginal cost of additional storage helps quantify the tradeoff between storing BAM files and regenerating them later. For projects with limited storage, the decision to store BAM files should be based on the expected number of times the data will be reanalyzed. If the data will be reanalyzed more than once, storing the BAM files is likely to be more cost-effective than regenerating them each time. ### Measuring Read Length Effects on Analysis Power The third measurement relates to how read length affects the information that can be extracted from sequencing data. A study on allele-specific expression analysis from single-nucleus RNA-seq data found that read length can increase the power to detect allele-specific expression, and that more information could be extracted from reads in intronic regions than exonic regions when using single-nucleus RNA-seq. The [allelic imbalance study](https://doi.org/10.1186/s13059-026-04062-6) demonstrated that experimental and computational choices impact the power of analysis methods. For long-read data management, this means that the read length distribution in your FASTQ files directly affects the value of the data for certain analyses, and this should be considered when deciding whether to retain raw data for future reanalysis with different methods. ## Records and Documentation Requirements Proper documentation of format decisions and data management practices is essential for reproducibility and for compliance with funding and journal requirements. The following records should be maintained for each project. ### Data Management Plan A data management plan should describe the formats that will be used for raw and processed data, the storage locations, the backup strategy, and the retention schedule. Many funding agencies require a data management plan as part of grant applications, and having a well-developed plan simplifies compliance. The plan should specify which files are considered the master copies and which are considered derived data. It should also specify the conditions under which derived data can be deleted and the process for regenerating it if needed. ### Analysis Log An analysis log records every analysis performed on the data, including the tools used, the parameters, and the input and output files. This log is essential for reproducing analyses and for understanding why results may differ between runs. The [Galaxy Training Network](https://training.galaxyproject.org/) provides guidance on creating reproducible analysis workflows, and many of the principles apply to documentation practices. Each analysis step should be recorded with enough detail that another researcher could repeat it without access to the original analyst. ### Version Control for Scripts and Parameters All analysis scripts and parameter files should be maintained under version control. This allows you to track changes over time and to reproduce analyses with the exact versions of tools and parameters that were used originally. The [nf-core documentation](https://nf-co.re/docs) describes how community pipelines use version control and containerization to ensure reproducibility, and these practices can be adapted for individual projects. Containerization ensures that the software environment is preserved, which is important for long-term reproducibility. ### Variant Interrogation Documentation For projects involving variant analysis, the documentation requirements extend to the variant interrogation process itself. A tutorial on variant interrogation in tumor samples presents a practical framework organized into four phases: planning, gathering resources, filtering and validation, and dissemination and storage. The [variant interrogation tutorial](https://doi.org/10.1371/journal.pcbi.1013924) emphasizes that the complexity of variant analysis pipelines, terminology, and tool selection remains a major barrier, especially for those new to the field or working in translational settings. The framework guides researchers through critical steps including assembling the tools, reference data, and variant annotation sets required for analysis, executing a systematic approach to prioritize meaningful variants, and ensuring findings are reproducible and accessible through transparent reporting and data sharing. These documentation practices apply directly to the format decision because the storage format determines whether the raw data needed for variant reanalysis remains available. ## Common Failure Patterns in Data Management Several common failure patterns lead to data loss or analysis problems. Recognizing these patterns can help you avoid them in your own projects. ### Failure Pattern One: Discarding FASTQ After Initial Analysis The most common failure pattern is discarding FASTQ files after the initial analysis is complete. This often happens when storage space is limited and the BAM files appear to contain all the necessary information. The problem emerges when a new reference genome is released or when a new analysis tool requires unaligned reads. The cost of this failure is the inability to perform new analyses without resequencing the samples. For rare or irreplaceable samples, this can be a catastrophic loss. The prevention is to treat FASTQ files as the archival master copy and to never delete them without explicit approval from the project lead. ### Failure Pattern Two: Inconsistent Alignment Parameters The second common failure pattern is using different alignment parameters for different samples in the same project. This can happen when samples are processed at different times or by different analysts. The result is that BAM files are not directly comparable, and downstream analyses may be biased by the alignment differences. The prevention is to document alignment parameters in a project-level configuration file and to use the same parameters for all samples unless there is a documented reason for variation. The [nf-core documentation](https://nf-co.re/docs) describes how pipeline configurations enforce consistent parameters across samples, and this approach can be adapted for individual projects. ### Failure Pattern Three: Insufficient Backup The third common failure pattern is relying on a single copy of data without adequate backup. This is particularly dangerous for FASTQ files, which are the archival master copy. If the primary storage fails and there is no backup, the data is lost permanently. The prevention is to maintain at least two copies of all archival data, preferably in different physical locations. Many institutions provide backup services for research data, and these should be used for all FASTQ files and other irreplaceable data. ### Failure Pattern Four: Not Planning for Data Growth The fourth common failure pattern is not planning for data growth. Long-read sequencing projects often generate more data than initially anticipated, and storage can fill up faster than expected. This leads to emergency decisions about which files to delete, which are often made without adequate consideration of the consequences. The prevention is to monitor storage usage regularly and to plan for data growth in advance. This includes estimating the storage requirements for planned sequencing runs and ensuring that sufficient storage is available before the data is generated. ### Failure Pattern Five: Ignoring Tool-Specific Format Requirements The fifth common failure pattern is assuming that all downstream tools accept the same input formats. Different tools have different requirements, and some tools that appear to accept BAM input may actually require specific alignment tags or quality metrics that are not present in all BAM files. This can lead to analysis failures or incorrect results that are difficult to diagnose. The prevention is to check the documentation for each tool before deciding on a storage strategy. The [Bioconductor project](https://bioconductor.org/) provides documentation for each package that specifies the required input formats, and the [EMBL-EBI training resources](https://www.ebi.ac.uk/training) provide guidance on tool selection and data format requirements. ## Limitations of Each Format Both FASTQ and BAM formats have limitations that affect their suitability for different purposes. Understanding these limitations helps researchers make informed decisions about data management. ### FASTQ Limitations FASTQ files do not contain any information about the sequencing platform or the basecalling model used to generate the reads. This information must be recorded separately in the metadata. For long-read platforms, the basecalling model can significantly affect read quality and error profiles, and this information is essential for interpreting the data. FASTQ files also do not contain information about read methylation or other base modifications that some long-read platforms can detect. If base modification information is needed, it must be stored in a separate format or in the platform-specific output files. ### BAM Limitations BAM files are reference-dependent, and the alignment information becomes outdated when new reference genomes are released. Realigning from FASTQ is the most reliable way to update alignments, but this requires access to the original FASTQ files. BAM files also do not preserve all information from the original reads. Soft-clipped bases are retained in the alignment, but the original read sequence may be truncated if the read extends beyond the reference genome. For reads that do not align to the reference, the BAM file may contain only the unmapped read sequence without any alignment information. ### Format Compatibility Considerations Not all downstream tools accept both formats. Some tools, particularly those designed for de novo assembly or reference-free analysis, require FASTQ input. Other tools, particularly variant callers and transcript quantification tools, require BAM input. Understanding the format requirements of your downstream tools is essential for making storage decisions. The [Bioconductor project](https://bioconductor.org/) provides a comprehensive ecosystem of tools for genomic analysis, and the documentation for each package specifies the required input formats. Checking the documentation for your planned tools before deciding on a storage strategy can prevent compatibility problems later. ## Quality and Welfare Controls in Data Management While the term welfare is typically associated with animal research, the principle of responsible data stewardship applies to all research data. The following controls help ensure that data is managed responsibly throughout its lifecycle. ### Data Integrity Checks Regular integrity checks ensure that stored files have not been corrupted. This is particularly important for archival data that may be stored for years. Checksums should be calculated when files are created and verified periodically. For FASTQ files, integrity checks should include verifying that the file can be read completely and that the sequence and quality lines are properly formatted. For BAM files, integrity checks should include verifying that the file can be read by standard tools and that the alignment information is internally consistent. ### Access Controls Access controls ensure that only authorized personnel can modify or delete archival data. This is particularly important for FASTQ files, which are the master copy. Access controls can be implemented at the file system level or through data management systems. For projects with multiple researchers, establish clear roles and responsibilities for data management. The project lead should have final authority over decisions to delete or modify archival data, and all changes should be documented. ### Retention Schedule A retention schedule specifies how long different types of data will be retained. This schedule should be based on funding requirements, journal policies, and the expected useful life of the data. For most research projects, FASTQ files should be retained for at least the duration of the funding period plus any required post-project retention period. The [NCBI data resources](https://www.ncbi.nlm.nih.gov/) provide long-term archival for published sequencing data, and depositing data in NCBI ensures that it will be available for the long term. For data that will not be deposited in a public repository, the retention schedule should be documented in the data management plan. ### Reproducibility Controls Reproducibility controls ensure that analyses can be repeated with the same results. This requires documenting also the input data formats but also the software environment, including tool versions and dependencies. The [nf-core documentation](https://nf-co.re/docs) describes how community pipelines use containerization to ensure that the software environment is preserved, and this approach can be adapted for individual projects. For long-read data, reproducibility also requires documenting the basecalling model and version, as these significantly affect read quality and error profiles. This information should be recorded in the analysis log alongside the alignment parameters. ## Safety and Regulatory Context Data management decisions have regulatory implications, particularly for projects involving human subjects or controlled data. The following considerations apply to these projects. ### Controlled Access Data For projects involving human genomic data, the FASTQ and BAM files may be subject to controlled access requirements. These requirements specify who can access the data and under what conditions. The storage and transfer of controlled access data must comply with institutional and regulatory requirements. The [NCBI data resources](https://www.ncbi.nlm.nih.gov/) provide controlled access repositories for human genomic data, and the submission process includes requirements for data use agreements and access controls. For projects that will generate controlled access data, the data management plan should specify how access will be managed. ### Data Transfer and Security The transfer of sequencing data between institutions or to public repositories requires attention to data security. For controlled access data, encrypted transfer methods should be used. For all data, the transfer process should be documented to ensure that files are not corrupted during transfer. The [EMBL-EBI training resources](https://www.ebi.ac.uk/training) provide guidance on data transfer and security best practices, and these principles apply to both FASTQ and BAM files. The format of the data does not change the security requirements, but the sensitivity of the data does. ### Publication and Sharing Requirements Many journals and funding agencies require that sequencing data be deposited in public repositories upon publication. The format of the deposited data is typically FASTQ for raw reads and processed files for derived data. Planning for data deposition early in the project simplifies compliance with these requirements. The [NCBI data resources](https://www.ncbi.nlm.nih.gov/) provide documentation on submission formats and requirements, and the submission process can be initiated before the project is complete. Depositing data early allows other researchers to access it and can facilitate collaborations. ### Data Provenance and Transparency Data provenance refers to the documented history of how data was generated, processed, and analyzed. For sequencing data, provenance includes the instrument settings, basecalling parameters, alignment parameters, and all subsequent analysis steps. Maintaining complete provenance is essential for regulatory compliance and for building trust in research findings. The [variant interrogation tutorial](https://doi.org/10.1371/journal.pcbi.1013924) emphasizes that ensuring findings are reproducible and accessible through transparent reporting and data sharing is a critical phase of the analysis workflow. This principle applies to format decisions because the choice of storage format affects what information can be preserved and shared. ## Professional Escalation Criteria Certain situations warrant escalation to institutional data management professionals or other experts. The following criteria indicate when professional assistance should be sought. ### Escalation Criterion One: Storage Exhaustion When storage capacity is exhausted or projected to be exhausted within a short timeframe, escalate to institutional data management services. They can provide guidance on storage options, data compression, and retention policies. Do not make unilateral decisions to delete data without consulting with data management professionals. ### Escalation Criterion Two: Data Corruption When data corruption is detected, escalate immediately to the appropriate technical support. Attempting to repair corrupted files without proper expertise can cause further damage. Preserve the corrupted files for analysis and do not overwrite them with potentially incomplete repairs. ### Escalation Criterion Three: Regulatory Compliance Questions When questions arise about regulatory compliance for data storage or sharing, escalate to the institutional review board or data privacy office. These questions are particularly important for human subjects data and for data subject to international transfer restrictions. ### Escalation Criterion Four: Major Format Migration When a major format migration is planned, such as moving from one reference genome to another or changing the primary analysis platform, escalate to bioinformatics support. They can provide guidance on the migration process and help avoid common pitfalls. ### Escalation Criterion Five: Workflow Performance Concerns When benchmarking reveals substantial performance variability between tools or workflows, escalate to bioinformatics support for guidance on workflow selection. The [benchmarking study](https://doi.org/10.3389/fbinf.2026.1803237) demonstrated that common tools can show substantial performance variability depending on the aligner chosen and the sequencing strategy used, and that aligner and primer design choices can significantly impact outcomes. Without systematic evaluation, researchers risk drawing inaccurate conclusions from suboptimal workflows. Bioinformatics support can help interpret benchmarking results and select appropriate tools for specific research questions. ## Frequently Asked Questions ### Should I store both FASTQ and BAM files for every project? Storing both formats is the safest approach for most research projects. FASTQ files preserve the raw data and allow realignment to new references, while BAM files support immediate analysis without the compute cost of realignment. The main exception is for projects with severe storage constraints where the cost of storing both formats is prohibitive. In those cases, prioritize FASTQ files because they can always be realigned to generate BAM files, but the reverse is not possible. ### How much storage space do long-read FASTQ files require? The storage requirement depends on the sequencing depth, read length distribution, and quality score encoding. A typical Nanopore run producing 10 to 20 gigabases of sequence data will generate FASTQ files in the range of 10 to 30 gigabytes before compression. Compression with gzip typically reduces the size by 3 to 4 fold. The exact size depends on the basecalling model and the quality score distribution, so measuring the actual file sizes for your platform is recommended. ### Can I convert BAM files back to FASTQ format? Yes, BAM files can be converted back to FASTQ format using standard bioinformatics tools. The conversion extracts the read sequences and quality scores from the BAM file and writes them in FASTQ format. However, the converted FASTQ file may not be identical to the original because some information may have been lost during alignment, such as the original read order or reads that were filtered during alignment. For archival purposes, the original FASTQ files are preferred. ### What is the best format for submitting data to public repositories? Most public repositories, including the NCBI Sequence Read Archive, accept FASTQ files as the primary format for raw sequencing data. FASTQ is the preferred format because it is platform-agnostic and does not depend on any particular reference genome. Some repositories also accept BAM files for aligned data, but FASTQ is the standard for raw read submission. Check the specific repository requirements before preparing your submission. ### How does the choice of reference genome affect BAM file storage? The reference genome version is a critical parameter for BAM files. BAM files aligned to one reference genome cannot be directly compared to BAM files aligned to a different reference genome. When a new reference genome is released, you must decide whether to realign your FASTQ files to the new reference or continue using the old reference for consistency. This decision should be documented in the data management plan. ### What quality metrics should I record for long-read data? Record the read length distribution, quality score distribution, total yield, and alignment rate for each sample. These metrics provide a baseline for evaluating data quality and for identifying samples that deviate from expected values. For long-read data, also record the basecalling model and version, as these significantly affect read quality and error profiles. ### How long should I retain FASTQ files after project completion? The retention period depends on funding requirements, journal policies, and the expected useful life of the data. For published data, the data should be retained for at least as long as the associated publication is expected to be cited. For data deposited in public repositories, the repository provides long-term archival. For data not deposited in a public repository, a retention period of at least 5 to 10 years is recommended. ### What should I do if I accidentally delete FASTQ files? If FASTQ files are accidentally deleted, check whether backups exist before taking any other action. If backups are available, restore the files from backup. If no backups exist, the data may be unrecoverable. In this case, document the loss and assess the impact on the project. If the data was deposited in a public repository, it may be possible to download the files from the repository. ## Related Bioinformatics Guides - [Genomic Data Analysis Tools: A Comparative Guide for Researchers](/knowledge/bioinformatics/genomic-data-analysis-tools-a-comparative-guide-for-researchers) - [Long-Read Metagenome Assembly: Overcoming Challenges with Nanopore and PacBio Data](/knowledge/bioinformatics/long-read-metagenome-assembly-overcoming-challenges-with-nanopore-and-pacbio-data) - [Multi-Omics Data Integration: A Comparative Framework for Choosing the Right Method](/knowledge/bioinformatics/multi-omics-data-integration-a-comparative-framework-for-choosing-the-right-method) - [Evaluating Metagenomic Assembly Tools: A Benchmarking Framework for Short-Read and Long-Read Data](/knowledge/bioinformatics/evaluating-metagenomic-assembly-tools-a-benchmarking-framework-for-short-read-and-long-read-data) - [Metabolomics Data Analysis in R: A Practical Workflow](/knowledge/bioinformatics/metabolomics-data-analysis-in-r-a-practical-workflow) ## References and Further Reading - [NCBI Data Resources](https://www.ncbi.nlm.nih.gov/). National Center for Biotechnology Information. - [EMBL-EBI Training](https://www.ebi.ac.uk/training). European Bioinformatics Institute. - [Bioconductor](https://bioconductor.org/). Bioconductor Project. - [Galaxy Training Network](https://training.galaxyproject.org/). Galaxy Project. - [nf-core Documentation](https://nf-co.re/docs). nf-core. - [The Carpentries Lessons](https://carpentries.org/lessons). The Carpentries. - [How benchmarking of bioinformatics tools is essential for informed workflow selection: a case study on SARS-CoV-2 subgenomic RNA detection.](https://doi.org/10.3389/fbinf.2026.1803237). 2026. - [Experimental and computational methods for allelic imbalance analysis from single-nucleus RNA-seq data.](https://doi.org/10.1186/s13059-026-04062-6). 2026. - [Tutorial for variant interrogation in tumor samples.](https://doi.org/10.1371/journal.pcbi.1013924). 2026. - [A comprehensive transcriptomic dataset of Sorghum bicolor seedlings under abiotic stress conditions.](https://doi.org/10.1038/s41597-026-07425-7). 2026. > This article is educational and does not replace validated analysis plans, institutional policy, clinical interpretation, or specialist review.