Human biomolecular data

Introduction

The study of human biomolecular data, including genomics, transcriptomics, proteomics, and metabolomics, has significantly advanced our understanding of the pathogenesis, treatment and prevention of infectious diseases. However, the quality of human biomolecular data can significantly impact the accuracy and reliability of the results obtained from analyses.

Quality control (QC) is an essential step in the analysis of human biomolecular data, particularly in infectious diseases. QC involves assessing the quality of the data generated during sample collection, processing, and analysis to identify and remove errors, biases, and artefacts that may affect the accuracy and interpretation of the results. In this page, we will discuss the essential QC steps for human biomolecular data specific to infectious diseases.

Sample collection and processing

The first step in QC for human biomolecular data is ensuring the safe proper collection and processing of samples. The Centers for Disease Control have published guidelines for safe handling of infectious agents, which can be found here. This includes collecting all clinical information that will be relevant for the analysis. For guidance and considerations related to the sample collection and processing specific to SARS-CoV, plese visit this documentation.

General Considerations

• Rapid and Efficient Sample Collection: During an infectious disease outbreak, it is crucial to collect samples as quickly and efficiently as possible to identify and contain the spread of the disease. A well-organised sampling plan should be implemented, including guidelines for sample collection, transport, and storage, and appropriate training should be provided to personnel.
• Safety Precautions: Safety precautions are crucial during an infectious disease outbreak to prevent the spread of the disease among the sample collection team and laboratory personnel. Appropriate protective equipment, such as gloves, masks, and gowns, should be provided to personnel, and guidelines for handling infectious samples should be strictly followed.
• Scalability: During an outbreak, the number of samples to be collected and processed may be significantly higher than normal. Therefore, the biomolecular analysis pipeline should be scalable and able to process a large number of samples quickly and accurately.
• High-Throughput Sample Processing: High-throughput sample processing methods, such as automated nucleic acid extraction and purification systems, can significantly increase the speed and efficiency of sample processing during an outbreak. These methods can help reduce the risk of contamination and human error and increase the accuracy and reliability of the results.
• Collaboration and Data Sharing: During an outbreak, collaboration and data sharing between different laboratories and research institutions can significantly accelerate the pace of research and help identify effective treatments and interventions quickly. Therefore, it is crucial to establish effective communication channels and data-sharing platforms to facilitate collaboration between different stakeholders.

Existing approaches

When looking for solutions to sample collection and processing, you can check this documentation.

Assessment of data quality and data preprocessing

Pre-analytical errors, such as wrong labelling of the specimen, the incorrect specimen collection device, incorrect specimen type are the most frequently reported sources of error (between 48% and 62%) within the clinical laboratory, as stated in this 2021 publication, Quality Assurance in the Clinical Virology Laboratory.

Pre-analytical testing (proficiency testing):

Biospecimen Proficiency Testing program is an external quality assessment tool to verify the precision and accuracy of biospecimen testing methods, and the efficiency of laboratory sample processing methods. At the end of the program, each participant will receive a performance report, a certificate of participation, and a label of participation (e.g. https://www.isber.org/page/PTGI/Proficiency-Testing-General-Information.htm). Proficiency testing schemes and providers can be found from e.g. https://www.eptis.org/.

Conducting quality control is an essential step to avoid these kinds of errors The use of a laboratory information management system (LIMS) streamlines the management of sample processing, data and laboratory workflows. These types of software increase the traceability of samples, ensuring data accuracy from sample collection to analysis. There are several commercially available LIMS systems and it is also possible to develop such software with open source and freely available solutions like REDCap (Vesteghem et al., 2022). Regardless of the LIMS software choice, analytical laboratories must always enforce the use of standardized procedures and audit trails, in order to maintain high levels of data integrity, compliance, and operational efficiency.

Positive and negative controls:

In analytical procedures it is important to select and use proper controls, such as negative or positive controls to ensure the accuracy and validity of the results. Positive controls are samples known to produce an expected outcome, ensuring that the analytical system is working property. Negative controls are samples that should not produce the result expected for positive samples. It helps establish the background noise or non-specific responses within the system. Laboratories should develop and validate a routine of analytical procedures using these types of controls with the aim of maintaining high quality results.

For samples that would be sequenced using NGS, some quality controls will assess sample mislabeling and contamination. For example, tools such as ​​CHARR will detect sample contamination. GRAF-sex will detect sample sex ensuring that it matches the labeled sex. PEDDY will infer family connections thus revealing any sample mixup during sample collection from families. Further details on these tools are outlined below.

It’s also important to notice the need for international standards, reference materials, and external quality assessment to support quality assurance. Although it can be a challenging and a complex process, verifying and validating assay methods, monitoring assay performance over time, and ensuring traceability to international standards for calibration are key points for high quality results.

Reference material:

The use of reference material is advised. Reference material is used to validate the quality and traceability of a sample and to validate the methods used to analyze these materials. Reference materials are often used in analytical chemistry to perform these validations, and most analytical instruments gather comparative data. Several standards are available from commercial vendors.

Reference material needs to be fit-for-purpose, i.e. it needs to address the nature of the study (clinical, research), and the technologies used in the analysis. To obtain credibility, reference material should be accepted by the scientific community, e.g. via publications, and public use. Self-made reference material which possibly cannot be shared across laboratories should not be used.

After assessing data quality, the next step is data preprocessing. Data preprocessing involves cleaning and filtering the data to remove errors, artifacts, and biases that may affect downstream analysis. Common preprocessing steps include read trimming, adapter removal, quality filtering, and read alignment. Tools such as Trimmomatic, Cutadapt, and BWA can be used for data preprocessing.

Considerations

• Sample Quality Control: The quality of the biological sample used for biomolecular analysis is critical for the reliability and accuracy of the resulting data. Therefore, it is essential to implement quality control measures to assess the quality of the sample, including its purity, integrity, and suitability for downstream analysis.
• Standardization and Calibration: Standardisation and calibration of the biomolecular analysis methods used are essential for ensuring the accuracy and reliability of the results. This includes the use of validated assays, appropriate positive and negative controls, and quality assessment metrics to monitor the performance of the assays.
• Reproducibility and Robustness: Reproducibility and robustness of the biomolecular analysis methods used are crucial for ensuring the consistency and reliability of the results across different laboratories and researchers. Therefore, it is essential to evaluate the reproducibility and robustness of the methods used, including the variability of results obtained by different operators, equipment, and reagents.
• Data Quality Assessment and Reporting: Data quality assessment and reporting are critical for ensuring the accuracy and reliability of the results and communicating the findings to stakeholders effectively. This includes implementing appropriate data quality control measures, such as quality assessment metrics and statistical analysis, to identify and correct potential sources of error and ensure the accuracy and reliability of the data. Additionally, clear and concise reporting of the results, including appropriate data visualisation and interpretation, can help communicate the findings effectively to stakeholders.
• Sequencing quality control: Information and tools for the assessment of the quality of the sequencing data in the context of COVID-19 can be found in the publication by the COVID-19 Host Genetics Initiative (HGI) on the Whole Exome/Genome Sequencing Analysis Plan.
• Benchmarking studies: To obtain comparative quality assessment of the applied laboratory and analytical procedure, the participation in benchmarking studies is advised. Benchmarking studies can be performed in different ways, depending on the scientific discipline. The quality of sample processing results of aliquots from the same sample across several laboratories can be compared and reasons for quality deviations can possibly be identified. Benchmarking bioinformatics workflows can be done e.g. by providing mock genomes to the study participants and identifying those workflows that perform best in in silico identification or discovery of the respective trait in question (e.g. a variant, deletion of certain length, etc.). In the field of COVID-19, several benchmarking studies have been performed, e.g. for the sequencing of viral genomes (Liu et al., 2021), or for benchmarking bioinformatics approaches for virus identification workflows (Wu et al., 2024); (Hegarty et al., 2023); (Hegarty et al., 2023); (de Vries et al.,).
• More broader, benchmarking studies for the identification of pathogens in human tissue have been published earlier (e.g. Gihawi et al., 2019; or antimicrobial resistance using sequencing Petrillo et al., 2021).

Human data removal from pathogen sequencing

When sequencing pathogen samples, generated data is often contaminated with host genetic material. Since a human genome is much larger than most pathogens, the sequencers usually provide much more human data than pathogen data.

Even traces of human genetic material in pathogen sequencing data could potentially identify individuals. The identifiability risk arises because human genome sequences are unique and classified as a special category of human data under GDPR. To ensure the ethical and legal handling of genomic data, removing human reads minimizes the risk of re-identification and ensures compliance with human data protection regulations.

Usually, the human reads are removed from pathogen sequencing data using two classes of methods: 1) ones that map all reads to the human genome (and discard those that do) or 2) ones that use a taxonomic classifier to predict the origin of each read.

Methods that use the first approach include CS_Score, DeconSeq, GenCoF and MetaGeniE. Methods that predict the origin of the reads include Centrifuge or Kraken.

It is noteworthy that both categories result in false negative calls - reads that are human but fail to be classified (or aligned) as such and, therefore, could be found in open repositories. Bush et al., 2020 were able to extract a sufficient number of reads to call known human SNPs in 6 % of the samples they analyzed from SRA BioProject. They developed a two-stage approach to remove human reads from pathogen sequencing experiments: first using Bowtie2 followed by SNAP. The benchmark was made in short read sequences from bacterial samples. Other types of sequencing (e.g. long read sequencing) might require other combinations of tools to provide pathogen data free from human reads.

It is important to check if the repository where pathogen data is being deposited recommends the use of a specific tool for quality control and to adhere to this prior or during the data upload step.

Existing approaches

Bioinformatics tools for quality control

The following bioinformatics tools are usually part of a bioinformatics pipeline that is used for the identification of a specific variant of a known pathogen. Therefore, they can be used for infectious diseases but also in other scientific domains.

1. Sequencing quality
   There are several features in a sequencing file (fastq) that indicate the quality of sequencing:
   • Base sequence quality
   • Base sequence content
   • Sequence GC content
   • Base N content
   • Sequence Duplication Levels
   • Overrepresented sequences
   • Adapter Content
   • Read length

   These metrics can be estimated using FastQC, which also provides graphic reports to help the user identify any issues.

2. Alignment quality
   Once the reads have been aligned to the respective reference genome, a number of additional quality metrics can be calculated for another layer of quality control. These include:
   • Insert size
   • Base coverage
   • Mapping quality
   • Percentage mapped
   • Number of uniquely mapped reads
   • Contamination

   Tools such as SAMtools and QualiMap can measure these metrics, and FastQ-Screen measures contamination with other species.

3. Variant call quality
   Upon identifying (calling) genetic variants using an appropriate tool (e.g., GATK), further quality checks need to be made:
   • Ratio transition/transversion (ti/tv) outside the expected range
   • Ratio het/hom
   • Call rate
   • Singletons**

   Tools vcftools, bcftools, and plink can calculate these metrics.

• FASTQC: FastQC is a widely used quality control tool for next-generation sequencing (NGS) data. It provides a comprehensive analysis of the sequence quality, including per base sequence quality, sequence length distribution, adapter content, and GC content. FastQC generates graphical reports that allow users to identify potential issues and assess the overall quality of the sequencing data.
• SAMtools: Samtools is a widely used suite of tools for working with sequencing data in the SAM/BAM format. It can be used for a variety of tasks, including sorting, merging, indexing, and filtering SAM/BAM files. Samtools also includes a quality control tool called “samtools flagstat,” which generates statistics on the number and proportion of reads that pass various quality filters, such as mapping quality, read length, and base quality.
• Bcftools: Bcftools is a set of tools for working with variant calls in the VCF format. It can be used for tasks such as filtering, merging, and comparing VCF files. Bcftools includes several quality control tools, such as “bcftools stats,” which generates summary statistics on the quality of variant calls, such as the number of variants called, the proportion of variants passing quality filters, and the distribution of variant quality scores.
• Qualimap: Qualimap is a quality control tool that assesses the quality of the sequencing data at different stages of the analysis pipeline, including read mapping, coverage, and expression analysis. It generates various graphical outputs that allow users to assess the quality and reliability of the data at each step of the analysis.
• Genome Analysis Toolkit (GATK): GATK is a widely used tool for variant calling and genotyping from NGS data. It uses various filtering and quality control options to identify high-quality variants and reduce false positives in the analysis.
• Picard: Picard is a suite of tools that provides quality control and processing of NGS data, including duplicate read removal, format conversion, and alignment. It is widely used in sequencing data analysis pipelines to improve the quality and reliability of the data.
• Trimmomatic: Trimmomatic is a tool used for the removal of adapter sequences, low-quality reads, and sequences with ambiguous bases from NGS data. It uses various filtering and trimming options to improve the quality of the sequencing data and prepare it for downstream analysis.
• FastQC Screen: FastQScreen is a quality control tool used to detect contamination in sequencing data (FASTQ files). It screens the input data against a database, formed by the genomes of all of the organisms that could have contaminated the sample, along with PhiX, Vectors or other contaminants commonly seen in sequencing experiments.
• GRAF pop: GRAF-pop is a software tool that infers the subject ancestry. For the estimation it uses about 100,000 SNPs from genotype datasets, which can be in PLINK and VCF format.
• GRAF sex: GRAF-sex determines subject sexes using the genotypes. It’s used to validate the self-reported sexes in phenotype datasets.

Diagnostic assays quality control

• Sample type, collection methods and transport to the laboratory: considering the test that is going to be performed while choosing the sample type and collection methods is a crucial step to ensure the validity and quality of the results. To get detailed guidelines for Collecting and Handling of Clinical Specimens for COVID-19 Testing, you can visit this Centers for Disease Control and Prevention documentation.
• Internal human control: an important aspect is human resource management, for instance, ensuring personnel training, competency, and continuous professional development is necessary.
• External quality assessment: external audits are an essential part for ensuring the quality of the procedures. They provide an objective overview that will state any discrepancies between the results and reveal points of improvement.
• Procedural documents: How an specific laboratory activity is conducted or how a piece of equipment is has been used must be recorded in standard operating procedures (SOPs) and equipment operating procedures (EOPs). These records will ensure results reproducibility. It’s also important to notice that these records must undergo regular reviews, ensuring that they remain compliant.
• Negative and positive controls: including negative controls it’s important to detect contamination in the reagents, consumables, and general laboratory environment. Including negative controls in each run of diagnostic molecular testing based on RT-PCR, for example, will avoid false positive results resulting from cross-contamination. As well as positive controls for the target sequence/s used in the individual assays, to ensure no false negative results due to reagents failure. For example, the study of a housekeeping human gene would ensure quality of the sample collection procedure, as it controls that the sample has human material and the quality of the analytical procedure, making it possible to check that there is no PCR inhibition and the reagents are working properly.

Collection of metadata

Proper metadata, adhering to the FAIR principles, significantly enhances the overall quality of biomedical studies. It enables researchers to find, access, and understand data, promotes interoperability for collaboration, and ensures that data is reusable for future research endeavours.

To ensure the correct and standard collection of metadata the definition of a metadata schema, beforehand, in an excel or spreadsheet, is a core decision. An interoperable, but yet, customizable metadata scheme can be found in this publication.

The mandatory and optional variables must be selected depending on the aim of the study and should include any relevant clinical data such as vaccination state, age, gender, or hospitalisation. More considerations about metadata can be found in the Data resources page.

While completing the forms, the use of ontologies is highly recommended:

• Infectious Disease Ontology (IDO), a suite of interoperable ontology modules aiming to provide coverage of all aspects of the infectious disease domain. Further details about IDO can be found in this paper.
• CovidO, an ontology for covid-19 metadata, is described in this paper.

Besides sample metadata, assay metadata describing how the results were generated is also necessary. This metadata will depend on the type of assay (antigenic, molecular, target genes of the assay, serological, etc) and it should include all the used variables on which the results depend on.

In conclusion, the meticulous collection of metadata stands as a cornerstone in the pursuit of conducting high-quality biomedical studies. Embracing and implementing robust metadata practices not only enhances the transparency, reproducibility, and reliability of research findings but also aligns seamlessly with the FAIR principles. By systematically documenting and organising essential contextual information, researchers contribute not only to the integrity of their own studies but also to the broader scientific community’s ability to harness and build upon valuable knowledge. Therefore, investing in comprehensive metadata collection is not just a procedural necessity; it is a commitment to elevating the overall quality and impact of biomolecular research.

Related pages

More information

Links to RDMkit

Skip tool table

Tools and resources on this page

Contributors