Stacking up RADSeq assembly programs: From complete hit to completely abysmal

Area of Science:

• Genomics
• Bioinformatics
• Molecular Ecology

Background:

• Decreasing sequencing costs fuel novel genotyping methods like reduced representation sequencing libraries (RRLs).
• RRLs, also known as genotyping by sequencing or restriction-site associated DNA sequencing, generate genome-wide data.
• Assembling RRL data de novo is challenging without a reference genome, with limited knowledge of program effectiveness.

Purpose of the Study:

• To systematically evaluate the performance of six common assembly programs for RRL data.
• To assess the impact of two key parameters on assembly completeness and accuracy.
• To provide recommendations for best practices in RRL data assembly and program evaluation.

Main Methods:

• Simulated double digests of Homo sapiens and Arabidopsis thaliana genomes were used.
• Varied mutation rates and types were incorporated into the simulated data.
• Six commonly used assembly programs and two parameters were tested for their effectiveness.

Main Results:

• Substantial variation in assembly performance was observed across different programs.
• The CD-HIT assembler demonstrated consistently high performance, though it is infrequently used.
• Several other tested assemblers failed to produce complete and accurate assemblies under various conditions.

Conclusions:

• The choice of assembly program significantly impacts the quality of RRL data analysis.
• CD-HIT is recommended as a highly effective assembler for RRL data.
• Standardized evaluation methods and best practices are crucial for future RRL assembly studies.