Whole-Genome Sequencing for Antimicrobial Resistance Prediction

By Dr. Zubair Khalid, DVM, MS, PhD · 2026-08-27

Quick Answer

• Whole-genome sequencing predicts antimicrobial resistance by detecting resistance genes and mutations in bacterial genomes, complementing phenotypic antimicrobial susceptibility testing.
• A practical workflow requires raw read quality control, read trimming, assembly or read-based analysis, and resistance gene detection using curated databases such as ResFinder or CARD.
• Interpretation must account for gene expression, promoter mutations, and novel resistance mechanisms, so WGS predictions cannot fully replace culture-based minimum inhibitory concentration testing.

At a Glance

Antimicrobial Resistance and the Role of Whole-Genome Sequencing

Antimicrobial resistance is the ability of bacteria to survive and grow in the presence of drugs that would normally inhibit or kill them. The World Health Organization identifies antimicrobial resistance as a global public health concern that threatens the effective treatment of infectious diseases and requires coordinated surveillance and stewardship efforts. Whole-genome sequencing has emerged as a powerful tool for characterizing bacterial pathogens because it provides the complete genetic content of an organism in a single experiment. For antimicrobial resistance prediction, WGS offers the potential to identify known resistance genes, detect mutations in chromosomal genes, and track the spread of resistant clones across clinical and environmental settings.

Traditional antimicrobial susceptibility testing measures the phenotypic response of a bacterium to a drug by determining the minimum inhibitory concentration, which is the lowest drug concentration that prevents visible growth. The Clinical and Laboratory Standards Institute publishes standards for antimicrobial susceptibility testing that define methods, quality control, and interpretive criteria for clinical laboratories. These standards-based methods remain the reference approach for determining whether a clinical isolate is susceptible, intermediate, or resistant to a specific drug. Whole-genome sequencing does not replace these methods but can complement them by providing genetic context for observed phenotypes and by identifying resistance mechanisms that may not be detected by phenotypic testing alone.

The practical problem for a bioinformatician or clinical microbiologist is how to convert raw sequencing data into a reliable prediction of antimicrobial resistance. This process involves multiple computational steps, each with its own sources of error and interpretation challenges. A well-designed workflow must account for sequencing platform differences, read quality, assembly quality, database completeness, and the biological complexity of resistance mechanisms. The remainder of this article describes a practical pipeline for using WGS data to predict antimicrobial resistance phenotypes, from raw reads to resistance gene identification and interpretation.

Core Principles of Genomic Resistance Prediction

Resistance Genes versus Resistance Mutations

Bacteria acquire antimicrobial resistance through two broad genetic mechanisms. The first is the acquisition of resistance genes carried on mobile genetic elements such as plasmids, transposons, and integrons. These genes encode enzymes that inactivate drugs, efflux pumps that expel drugs, or proteins that protect the drug target. Examples include beta-lactamase genes that hydrolyze beta-lactam antibiotics, aminoglycoside-modifying enzymes, and tetracycline efflux pumps. The second mechanism is the mutation of chromosomal genes that encode drug targets or regulatory proteins. Mutations in genes such as gyrA and parC can confer resistance to fluoroquinolones, while mutations in the promoter region of the bla genes can increase beta-lactamase expression.

WGS-based resistance prediction works best when the resistance mechanism is a known gene with a clear association between gene presence and resistance phenotype. For acquired resistance genes, the presence of the gene in the genome is usually sufficient to predict resistance to the corresponding drug class. For chromosomal mutations, the prediction requires knowledge of specific mutations and their effect on the drug target. Some mutations are well characterized, while others have uncertain effects. The accuracy of the prediction depends on the completeness of the reference database and the quality of the sequence data.

Genotype-Phenotype Concordance

The relationship between the presence of a resistance gene and the observed phenotype is not always perfect. A gene may be present but not expressed, or it may be expressed at a level too low to produce a resistant phenotype. Conversely, a resistance phenotype may arise from a mechanism that is not captured by the resistance gene database. The concordance between genotype and phenotype varies by species, drug class, and the specific resistance mechanism. For example, the presence of a beta-lactamase gene in Escherichia coli is strongly associated with resistance to ampicillin, but the same gene may not confer resistance to a cephalosporin if the enzyme has a narrow substrate profile.

The interpretation of WGS results must therefore be performed in the context of the organism and the drug being tested. A resistance gene that is known to confer resistance to a specific drug class should be reported as a predictor of resistance, but the final clinical decision should be based on the phenotypic antimicrobial susceptibility testing result. The Clinical and Laboratory Standards Institute standards provide the framework for interpreting antimicrobial susceptibility testing results and for determining the clinical relevance of a resistance phenotype. WGS data can inform the selection of drugs for testing and can explain unexpected phenotypic results, but it does not replace the phenotypic test.

Practical Workflow for WGS-Based Resistance Prediction

Step 1: Sample Preparation and DNA Extraction

The first step in the WGS workflow is the preparation of a pure bacterial culture and the extraction of high-quality genomic DNA. The quality of the DNA is critical because it directly affects the quality of the sequencing data. The American Type Culture Collection provides authenticated bacterial strains that can be used as controls for sequencing and resistance gene detection. These reference strains have known resistance profiles and can be used to validate the performance of the sequencing and analysis pipeline.

DNA extraction should be performed using a method that produces high-molecular-weight DNA with minimal contamination from proteins, polysaccharides, and other cellular components. The quantity and quality of the DNA should be assessed using a fluorometric method and a spectrophotometric method. The DNA concentration should be sufficient for the library preparation protocol, and the purity should be assessed by the A260/A280 ratio. The DNA should be stored at the appropriate temperature to prevent degradation.

Step 2: Sequencing Platform Selection

The choice of sequencing platform affects the read length, the error rate, and the cost of the sequencing run. Short-read platforms such as Illumina produce high-quality reads of 150 base pairs or longer, with a low error rate. These platforms are well suited for resistance gene detection because the genes are typically short and can be assembled from short reads. Long-read platforms such as Oxford Nanopore and Pacific Biosciences produce reads that are tens of thousands of base pairs long, which can resolve repetitive regions and provide complete genome assemblies. However, long-read platforms have a higher error rate, which can affect the accuracy of resistance gene detection.

The choice of platform should be based on the research question and the available resources. For routine resistance gene detection, short-read sequencing is often sufficient. For the assembly of complete genomes or the detection of resistance genes in repetitive regions, long-read sequencing may be necessary. The sequencing depth should be sufficient to cover the genome at least 30 times for short-read data and at least 20 times for long-read data. The depth of coverage affects the confidence in the detection of resistance genes, especially for genes that are present in low copy number.

Step 3: Quality Control of Raw Reads

The first computational step in the WGS workflow is the quality control of the raw sequencing reads. The raw reads are stored in FASTQ files that contain the sequence and a quality score for each base. The quality scores are used to identify low-quality bases and to assess the overall quality of the sequencing run. The FastQC tool generates a report that includes the per-base quality score, the GC content, the sequence length distribution, and the level of adapter contamination. The MultiQC tool can aggregate the FastQC reports from multiple samples into a single report for comparison.

The quality control step is essential because low-quality reads can lead to false positive resistance gene detection. A low-quality base can be misread as a resistance gene variant, or a short read can be incorrectly aligned to a resistance gene. The quality control report should be reviewed to identify any samples that fail the quality thresholds. Samples with a high proportion of low-quality bases or adapter contamination should be re-sequenced or the reads should be trimmed.

Step 4: Read Trimming

Read trimming is the process of removing low-quality bases and adapter sequences from the reads. The trimming step is performed after the quality control report has been reviewed. The Trimmomatic and fastp tools are commonly used for read trimming. The trimming parameters should be set to remove bases with a quality score below a threshold, typically 20, and to remove adapter sequences that are present at the ends of the reads. The trimming step should be performed with caution because over-trimming can remove short resistance genes or the ends of genes that are important for the detection.

The trimmed reads should be re-analyzed with FastQC to confirm that the trimming has improved the quality. The trimming step should be documented in the analysis log, including the number of reads before and after trimming and the parameters used. The trimmed reads are then used for the assembly or the read-centric analysis.

Step 5: Genome Assembly or Read-Centric Analysis

The next step is to assemble the trimmed reads into a genome or to map the reads to a reference genome. The assembly step is performed using a de novo assembler such as SPAdes, which generates a set of contiguous sequences called contigs. The assembly quality should be assessed by the number of contigs, the N50 value, and the total length of the assembly. A high-quality assembly should have a small number of contigs and a high N50 value. The assembly can be used for the detection of resistance genes, but the assembly process can introduce errors, especially in repetitive regions.

The read-centric analysis is an alternative to the assembly. In this approach, the trimmed reads are mapped to a reference genome or to a database of resistance genes. The mapping is performed using a read aligner such as BWA. The mapped reads are then used to identify the presence of resistance genes. The read-centric analysis is faster than the assembly and can be used for the detection of resistance genes in a large number of samples. However, the read-centric analysis is limited by the completeness of the reference database and the quality of the mapping.

Step 6: Resistance Gene Detection

The resistance gene detection is performed using a tool that compares the assembled contigs or the mapped reads against a database of resistance genes. The ResFinder database is a curated collection of resistance genes that are associated with antimicrobial resistance. The CARD database is a comprehensive database of resistance genes and the associated mutations. The ABRicate tool can be used to screen the contigs against multiple databases, including ResFinder and CARD.

The detection tool should be run with the appropriate parameters, including the minimum identity and the minimum coverage. The minimum identity is the percentage of the gene sequence that must match the database sequence, and the minimum coverage is the percentage of the gene sequence that must be present in the sample. A common threshold is 90 percent identity and 60 percent coverage, but the thresholds should be adjusted based on the database and the research question. The output of the detection tool is a table that lists the resistance genes that are present in the sample, the percentage identity, and the percentage coverage.

Step 7: Interpretation and Reporting

The interpretation of the resistance gene detection results requires the integration of the genomic data with the phenotypic data. The presence of a resistance gene is a predictor of resistance, but the prediction must be confirmed by the antimicrobial susceptibility testing. The interpretation should consider the following factors:

• The identity and coverage of the resistance gene
• The known function of the resistance gene
• The drug class that the gene confers resistance to
• The presence of other resistance genes that may interact
• The quality of the assembly and the sequencing data

The reporting of the WGS results should include the list of resistance genes, the identity and coverage values, and the predicted resistance phenotype. The report should also include the limitations of the analysis, including the possibility of false positives and false negatives. The report should be reviewed by a clinical microbiologist or a bioinformatician with experience in antimicrobial resistance.

Tools and Databases for Resistance Gene Detection

ResFinder

ResFinder is a web-based tool that identifies acquired resistance genes in whole-genome sequencing data. The tool uses a database of known resistance genes and a BLAST-based search to identify the genes in the input data. The input can be a set of contigs or a set of raw reads. The output is a table that lists the resistance genes, the identity, the coverage, and the predicted resistance phenotype. ResFinder is a widely used tool for the detection of acquired resistance genes, and it is maintained by the Center for Genomic Epidemiology.

The ResFinder database is curated and updated regularly to include new resistance genes. The database includes genes for resistance to beta-lactams, aminoglycosides, tetracyclines, fluoroquinolones, and other drug classes. The tool is available as a web server and as a command-line tool. The web server is easy to use and is suitable for the analysis of a small number of samples. The command-line tool is suitable for the analysis of a large number of samples and can be integrated into a bioinformatics pipeline.

CARD

The Comprehensive Antibiotic Resistance Database (CARD) is a database of resistance genes and the associated mutations. The database is curated and includes the resistance genes, the drug classes, and the resistance mechanisms. The CARD database is used by the Resistance Gene Identifier (RGI) tool, which is a tool for the detection of resistance genes in a genome. The RGI tool uses a hidden Markov model to identify the resistance genes in the input. The output of the RGI tool is a table of the resistance genes with the identity, the coverage, and the predicted resistance.

The CARD database is a comprehensive database that includes the resistance genes and the mutations. The database is updated regularly to include new resistance genes and mutations. The RGI tool is available as a web server and as a command-line tool. The web server is suitable for the analysis of a small number of samples, and the command-line tool is suitable for the analysis of a large number of samples.

ABRicate

ABRicate is a tool that screens a genome assembly against multiple resistance databases. The tool is a command-line tool that can be used to screen the assembly against the ResFinder, CARD, and other databases. The tool uses a BLAST-based approach to identify the resistance genes in the assembly. The output of the tool is a table of the resistance genes with the identity, the coverage, and the predicted resistance.

ABRicate is a useful tool for the comparison of the results from multiple databases. The tool can be used to identify the resistance genes that are detected by multiple databases and to identify the genes that are detected by only one database. The tool is also useful for the detection of resistance genes in a large number of samples, as it can be run in a batch mode.

Quality Control and Validation

The quality control of the WGS pipeline is essential for the reliability of the resistance prediction. The quality control should be performed at each step of the pipeline, from the DNA extraction to the resistance gene detection. The quality control should include the following:

• The quality of the DNA extraction
• The quality of the sequencing run
• The quality of the assembly
• The quality of the resistance gene detection

The quality of the DNA extraction should be assessed by the quantity and the purity of the DNA. The quality of the sequencing run should be assessed by the quality score of the reads and the depth of coverage. The quality of the assembly should be assessed by the number of contigs and the N50 value. The quality of the resistance gene detection should be assessed by the identity and the coverage of the detected genes.

The validation of the pipeline should be performed using a set of reference strains with known resistance profiles. The reference strains can be obtained from the American Type Culture Collection. The validation should include the detection of the known resistance genes and the prediction of the resistance phenotype. The validation should be performed on a regular basis to ensure that the pipeline is working correctly.

Common Failure Patterns and Troubleshooting

The WGS pipeline can fail at any step, and the failure can lead to incorrect resistance predictions. The common failure patterns include the following:

• Low-quality reads
• Adapter contamination
• Poor assembly
• False positive resistance gene detection
• False negative resistance gene detection

The low-quality reads can be caused by a poor sequencing run or a poor DNA extraction. The low-quality reads should be trimmed or the sample should be re-sequenced. The adapter contamination can be caused by the adapter sequences that are not removed during the library preparation. The adapter contamination should be removed during the trimming step.

The poor assembly can be caused by the repetitive regions of the genome or the low depth of coverage. The poor assembly can be improved by increasing the depth of coverage or by using a long-read sequencing platform. The false positive resistance gene detection can be caused by the presence of a gene that is not expressed or by the presence of a gene that is not associated with resistance. The false positive detection can be reduced by using a higher identity threshold and a higher coverage threshold.

The false negative resistance gene detection can be caused by the absence of the gene in the database or by the presence of a gene that is not detected by the tool. The false negative detection can be reduced by using a more comprehensive database and by using a lower identity threshold and a lower coverage threshold.

Limitations of WGS-Based Resistance Prediction

The WGS-based resistance prediction has several limitations that should be considered when interpreting the results. The limitations include the following:

• The prediction is based on the presence of known resistance genes
• The prediction does not account for the expression of the resistance genes
• The prediction does not account for the novel resistance mechanisms
• The prediction is not a substitute for the antimicrobial susceptibility testing

The prediction is based on the presence of known resistance genes, and the prediction is limited by the completeness of the database. The database may not include all the resistance genes that are present in the sample. The prediction does not account for the expression of the resistance genes. A gene that is present in the genome may not be expressed, and the gene may not confer resistance. The prediction does not account for the novel resistance mechanisms, and the prediction may be incorrect for a novel mechanism.

The prediction is not a substitute for the antimicrobial susceptibility testing. The antimicrobial susceptibility testing is the gold standard for the determination of the resistance phenotype. The WGS-based prediction should be used as a complement to the antimicrobial susceptibility testing, and the clinical decision should be based on the antimicrobial susceptibility testing.

Surveillance and Public Health Context

The WGS-based resistance prediction is a valuable tool for the surveillance of antimicrobial resistance. The World Health Organization identifies antimicrobial resistance as a global public health concern and recommends the surveillance of the resistance to inform the stewardship and the control of the resistance. The WGS-based prediction can be used to identify the resistance genes that are present in the bacterial population and to track the spread of the resistance genes across the population.

The surveillance of the antimicrobial resistance is performed by the collection of the bacterial isolates and the testing of the resistance. The WGS-based prediction can be used to identify the resistance genes in the isolates and to compare the resistance genes across the isolates. The surveillance can be used to identify the emergence of the new resistance genes and to inform the development of the new drugs.

The WGS-based prediction can also be used to identify the resistance genes in the food and the environment. The Bacteriological Analytical Manual provides the official methods for the detection of the foodborne pathogens. The WGS-based prediction can be used to identify the resistance genes in the foodborne pathogens and to inform the control of the resistance.

Professional Escalation Criteria

The WGS-based resistance prediction should be escalated to a clinical microbiologist or a bioinformatician with experience in the microbiology when the following conditions are met:

• The prediction is inconsistent with the antimicrobial susceptibility testing
• The prediction is based on a gene that is not in the database
• The prediction is based on a gene that is not expressed
• The prediction is based on a novel resistance mechanism

The escalation should be performed to ensure that the prediction is correct and that the clinical decision is based on the correct information. The escalation should be performed in a timely manner to avoid the delay in the treatment.

Decision Framework for Reporting WGS Resistance Predictions

A resistance gene detection result is not a final answer. The output from ResFinder, CARD, or ABRicate provides a candidate list of genetic determinants, but the transition from that list to a defensible resistance prediction requires a structured decision process. This section presents a practical framework for evaluating detection results, assigning confidence levels, and deciding when a prediction is actionable or when it must be deferred to phenotypic testing.

Confidence Tier Assignment

The first step in the decision framework is to assign each detected gene to a confidence tier based on the identity, coverage, and the biological plausibility of the gene-phenotype association. The tier determines whether the result can be reported as a prediction or whether it requires additional investigation.

Tier 1 includes genes with high identity and high coverage against a curated database entry, where the gene has a well-established association with a resistance phenotype in the species being analyzed. A beta-lactamase gene such as blaTEM in Escherichia coli with 100 percent identity and 100 percent coverage is a Tier 1 result for ampicillin resistance. Tier 1 results can be reported as strong predictors of resistance, but they still require confirmation by phenotypic testing before a clinical decision is made.

Tier 2 includes genes with moderate identity or coverage, or genes whose association with a resistance phenotype is less well established. A gene with 95 percent identity and 80 percent coverage may represent a novel variant of a known resistance gene, and the variant may have a different substrate profile than the reference gene. Tier 2 results should be reported as possible predictors of resistance, with a clear statement that the prediction is uncertain.

Tier 3 includes genes with low identity or coverage, or genes that are not known to confer resistance in the species being tested. A gene with 85 percent identity and 50 percent coverage may be a partial match to a resistance gene, but the match may not represent a functional resistance determinant. Tier 3 results should not be reported as resistance predictions. They should be recorded in the analysis log and flagged for further investigation.

The thresholds for the tiers should be defined before the analysis begins and should be documented in the standard operating procedure. The thresholds will vary by database and by the intended use of the results. A surveillance study may use a lower threshold to capture all potential resistance genes, while a clinical report may use a higher threshold to reduce the risk of false positives.

Gene Function and Expression Context

The presence of a resistance gene does not guarantee that the gene is expressed or that the expression level is sufficient to produce a resistant phenotype. The decision framework must account for the biological context of the gene, including the promoter region, the copy number, and the regulatory environment.

Some resistance genes are constitutively expressed, while others are inducible. An inducible gene may be present in the genome but not expressed during the antimicrobial susceptibility testing, resulting in a susceptible phenotype despite the presence of the gene. The framework should include a check for known promoter mutations that increase gene expression. For example, mutations in the promoter region of the bla genes can increase beta-lactamase production and change the resistance phenotype from susceptible to resistant.

The copy number of a resistance gene can also affect the phenotype. A gene present on a multicopy plasmid may be expressed at a higher level than a gene present in a single copy on the chromosome. The read depth from the sequencing data can provide an estimate of the copy number. A gene with a read depth that is significantly higher than the average genome depth may be present in multiple copies, and the prediction should account for the increased expression.

The decision framework should include a step for reviewing the genomic context of each detected gene. The contig or the read alignment should be examined to determine whether the gene is complete, whether the promoter region is present, and whether the gene is located on a mobile genetic element. The presence of a mobile genetic element can increase the risk of horizontal gene transfer and the spread of the resistance gene to other bacteria.

Species-Specific Interpretation

The same gene can have different effects in different species. A gene that confers resistance to a drug in one species may not confer resistance in another species because of differences in the drug target, the permeability of the cell wall, or the efflux systems. The decision framework must include species-specific interpretation rules.

The species identification should be confirmed before the resistance prediction is made. The species can be identified from the WGS data using a tool such as the average nucleotide identity or the 16S rRNA gene sequence. The species identification should be consistent with the phenotypic identification performed in the laboratory. If the species identification is uncertain, the resistance prediction should be deferred until the species is confirmed.

The species-specific interpretation rules should be based on the published literature and the curated databases. The CARD database includes information about the species-specific effects of the resistance genes. The ResFinder database includes the drug classes that are affected by each gene. The interpretation rules should be documented in the standard operating procedure and reviewed regularly as new information becomes available.

Reporting Categories

The decision framework should define the reporting categories for the resistance prediction. The categories should be consistent with the terminology used in the antimicrobial susceptibility testing. The categories are susceptible, resistant, and uncertain.

A susceptible prediction is made when no resistance genes or mutations are detected, and the genome does not contain any known resistance mechanisms. The susceptible prediction is only valid if the database is complete for the species and the drug class. A susceptible prediction should be reported with the caveat that novel resistance mechanisms may not be detected.

A resistant prediction is made when a Tier 1 gene is detected, and the gene is known to confer resistance to the drug class in the species. The resistant prediction should be reported with the gene name, the identity, the coverage, and the drug class. The resistant prediction should be confirmed by the antimicrobial susceptibility testing before the clinical decision is made.

An intermediate prediction is made when a Tier 2 gene is detected, or when the gene is present but the expression is uncertain. The intermediate prediction should be reported as a possible resistance, and the antimicrobial susceptibility testing should be performed to confirm the phenotype.

The reporting categories should be defined in the standard operating procedure, and the report should include the category for each drug class. The report should also include the limitations of the prediction, including the possibility of false positives and false negatives.

Escalation Criteria

The decision framework should include clear escalation criteria for the results that cannot be interpreted with confidence. The escalation criteria should be defined in the standard operating procedure and should be followed consistently.

The results should be escalated when the WGS prediction is inconsistent with the antimicrobial susceptibility testing. The inconsistency may be caused by a gene that is not expressed, a gene that is not in the database, or a novel resistance mechanism. The inconsistency should be investigated by a clinical microbiologist or a bioinformatician with experience in antimicrobial resistance.

The results should be escalated when the gene is not in the database. The gene may be a novel resistance gene that has not been characterized. The gene should be analyzed to determine the function and the association with the resistance phenotype. The gene should be reported to the database curator for inclusion in the database.

The results should be escalated when the gene is present but the expression is uncertain. The gene may be present in the genome but not expressed, or the gene may be expressed at a level that is not sufficient to produce a resistant phenotype. The expression should be confirmed by a phenotypic test or by a gene expression analysis.

The results should be escalated when the species identification is uncertain. The species identification is critical for the interpretation of the resistance genes. The species should be confirmed by a phenotypic test or by a molecular test.

Record System for Decision Documentation

The decision framework should be documented in a record system that captures the analysis steps, the decision criteria, and the final prediction. The record system should be designed to support the audit and the review of the results.

The record should include the sample identifier, the species identification, the sequencing platform, the read quality, the assembly quality, and the resistance gene detection results. The record should include the decision framework that was applied, the confidence tier for each gene, and the final prediction for each drug class.

The record should also include the antimicrobial susceptibility testing results, the comparison of the WGS prediction to the phenotypic result, and the resolution of any inconsistencies. The record should be reviewed by a clinical microbiologist or a bioinformatician before the final report is issued.

The record system should be implemented in a laboratory information management system or a spreadsheet. The record system should be validated to ensure that the data are accurate and complete. The record system should be backed up regularly to prevent the loss of the data.

Implementation Steps

The decision framework should be implemented in the laboratory with the following steps:

1. Define the confidence tiers and the thresholds for the identity and the coverage. The thresholds should be based on the database and the intended use of the results.
2. Define the reporting categories and the criteria for each category. The categories should be based on the antimicrobial susceptibility testing terminology.
3. Define the escalation criteria and the process for the escalation. The escalation process should be documented in the standard operating procedure.
4. Implement the record system and the documentation requirements. The record system should be validated and the staff should be trained on the use of the system.
5. Validate the decision framework using a set of reference strains with known resistance profiles. The reference strains can be obtained from the American Type Culture Collection. The validation should include the detection of the known resistance genes and the prediction of the resistance phenotype.
6. Review the decision framework on a regular basis and update the framework as new information becomes available. The review should include the new resistance genes, the new mutations, and the new species-specific effects.

Common Failure Patterns in Decision Making

The decision framework can fail when the criteria are not applied consistently or when the data is not interpreted correctly. The common failure patterns include the following:

• The use of a single threshold for all species and drug classes. The threshold should be adjusted based on the species and the drug class.
• The failure to account for the gene expression. The gene may be present but not expressed, and the prediction may be incorrect.
• The failure to account for the gene copy number. The gene may be present in multiple copies, and the prediction may be incorrect.
• The failure to confirm the species identification. The species identification is critical for the interpretation of the resistance genes.
• The failure to document the decision process. The decision process should be documented to support the audit and the review.

The failure patterns can be reduced by the training of the staff and the validation of the decision framework. The staff should be trained on the interpretation of the resistance genes and the application of the decision framework. The validation should be performed on a regular basis to ensure that the framework is working correctly.

Biosafety and Containment Context

The decision framework should be implemented in the context of the laboratory biosafety and containment. The laboratory should follow the biosafety guidelines for the handling of the bacterial isolates and the sequencing data. The biosafety guidelines are provided by the Biosafety in Microbiological and Biomedical Laboratories, which is published by the CDC and the NIH. The guidelines define the containment levels for the different types of the microorganisms and the practices for the safe handling of the microorganisms.

The sequencing data should be handled in a secure environment to prevent the unauthorized access. The data should be stored on a secure server and the access should be restricted to the authorized personnel. The data should be backed up to prevent the loss of the data.

The laboratory should follow the professional microbiology practice guidelines provided by the American Society for Microbiology. The guidelines provide the standards for the laboratory practice, the quality control, and the professional development. The laboratory should participate in the proficiency testing programs to ensure the quality of the results.

Professional Escalation Criteria

The decision framework should include the professional escalation criteria for the results that require the expertise of a clinical microbiologist or a bioinformatician. The escalation criteria are the following:

• The WGS prediction is inconsistent with the antimicrobial susceptibility testing.
• The gene is not in the database.
• The gene is present but the expression is uncertain.
• The species identification is uncertain.
• The resistance mechanism is novel.

The escalation should be performed in a timely manner to avoid the delay in the treatment. The escalation should be documented in the record system and the outcome of the escalation should be recorded.

Frequently Asked Questions

What is the difference between whole-genome sequencing and antimicrobial susceptibility testing?

Whole-genome sequencing identifies the genetic determinants of resistance, while antimicrobial susceptibility testing measures the phenotypic response of the bacterium to a drug. WGS can predict resistance based on the presence of known resistance genes, but the antimicrobial susceptibility testing is the gold standard for the determination of the resistance phenotype.

How long does it take to get a WGS-based resistance prediction?

The time required for the WGS-based resistance prediction depends on the sequencing platform and the analysis pipeline. The sequencing can take from a few hours to a few days, and the analysis can take from a few hours to a few days. The total time is typically longer than the time for the antimicrobial susceptibility testing.

What is the minimum depth of coverage for the WGS-based resistance prediction?

The minimum depth of coverage is the number of times the genome is sequenced. The minimum depth of coverage is typically 30 times for the short-read sequencing and 20 times for the long-read sequencing. The depth of coverage affects the confidence in the detection of the resistance genes.

What is the minimum identity and coverage threshold for the resistance gene detection?

The minimum identity is the percentage of the gene sequence that matches the database sequence, and the minimum coverage is the percentage of the gene sequence that is present in the sample. The common threshold is 90 percent identity and 60 percent coverage, but the threshold should be adjusted based on the database and the research question.

Can WGS detect all resistance mechanisms?

WGS can detect the resistance mechanisms that are known and that are in the database. WGS cannot detect the resistance mechanisms that are novel or that are not in the database. The prediction is limited by the completeness of the database.

How should the WGS-based resistance prediction be reported?

The report should include the list of the resistance genes, the identity and the coverage, and the predicted resistance phenotype. The report should also include the limitations of the prediction, including the false positives and the false negatives. The report should be reviewed by a clinical microbiologist or a bioinformatician.

What are the common failure patterns in the WGS-based resistance prediction?

The common failure patterns include the low-quality reads, the adapter contamination, the poor assembly, the false positive detection, and the false negative detection. The failure patterns can be reduced by the quality control and the validation of the pipeline.

What is the role of the WGS-based resistance prediction in the surveillance of antimicrobial resistance?

The WGS-based resistance prediction is a tool for the surveillance of antimicrobial resistance. The prediction can be used to identify the resistance genes in the bacterial population and to track the spread of the resistance. The surveillance can be used to inform the treatment and the control of the resistance.

Using the Evidence

Related Bacteriology Guides

• Biological Foundation Models for Antimicrobial Peptide Discovery in Veterinary Pathogens
• Aeromonas hydrophila in Aquaculture: Virulence Factors and Antimicrobial Resistance
• Common Bacterial Pathogens in Chickens: Identification and Antimicrobial Management
• Swine Bacterial Diseases: Colibacillosis, UTI, and Antimicrobial Resistance
• Avian Enrofloxacin: Pharmacology, Clinical Use, and Resistance in Poultry

References and Further Reading

• Biosafety in Microbiological and Biomedical Laboratories. CDC and NIH.
• Microbiology Resources. American Society for Microbiology.
• Antimicrobial Resistance. World Health Organization.
• Bacteriological Analytical Manual. U.S. Food and Drug Administration.
• Antimicrobial Susceptibility Testing. Clinical and Laboratory Standards Institute.
• Culture Collections. American Type Culture Collection.
• NCBI Pathogen Detection. National Center for Biotechnology Information.
• Whole-genome sequencing of Mycobacterium tuberculosis for prediction of drug resistance.. Epidemiology and infection, 2022.
• Whole-Genome Sequencing for Drug Resistance Profile Prediction in Mycobacterium tuberculosis.. Antimicrobial agents and chemotherapy, 2019.
• Whole genome sequencing and antimicrobial resistance among clinical isolates of Shigella sonnei in Addis Ababa, Ethiopia.. PloS one, 2024.
• Whole-genome sequencing, strain composition, and predicted antimicrobial resistance of Streptococcus pneumoniae causing invasive disease in England in 2017-20: a prospective national surveillance study.. The Lancet. Microbe, 2025.

This article is educational and does not replace validated laboratory procedures, clinical diagnosis, institutional biosafety review, or public-health guidance.