Systematic evaluation of de novo mutation calling tools using whole genome sequencing data

Area of Science:

• Genetics
• Bioinformatics
• Genomic Medicine

Background:

• De novo mutations (DNMs) are novel genetic alterations in offspring, frequently linked to severe developmental disorders.
• Accurate identification of DNMs is critical, especially with the increasing use of next-generation sequencing (NGS).
• Numerous bioinformatics tools exist for DNM detection, but a systematic comparison is lacking.

Purpose of the Study:

• To systematically evaluate and compare the performance of five de novo mutation (DNM) calling tools.
• To assess the concordance and accuracy of these tools using both real and simulated whole genome sequencing (WGS) trio data.
• To provide evidence-based recommendations for selecting appropriate DNM callers for WGS trio analysis.

Main Methods:

• Utilized real whole genome sequencing (WGS) trio data from the 1000 Genomes Project (1000G).
• Employed an in-house simulated trio dataset spiked with 100 known de novo mutations (DNMs).
• Evaluated five DNM calling tools: DeNovoGear, TrioDeNovo, PhaseByTransmission, VarScan 2, and DeNovoCNN.

Main Results:

• Observed low concordance rates: 8.4% for real data and 3.9% for simulated data, with most DNMs identified by only one tool.
• DeNovoGear demonstrated the highest F1 score on the real 1000G dataset.
• DeNovoCNN achieved the highest F1 score on the simulated WGS dataset.

Conclusions:

• Current de novo mutation (DNM) calling tools exhibit limited concordance on whole genome sequencing (WGS) trio data.
• Tool performance varies between real and simulated datasets, highlighting the need for careful selection.
• This comparative analysis offers valuable guidance for researchers and clinicians in choosing and applying DNM callers.