De novo likelihood-based measures for comparing genome assemblies.

Area of Science:

• Genomics
• Bioinformatics
• Computational Biology

Background:

• Genome assemblers are crucial for reconstructing genomes from sequencing reads.
• Current quality assessment relies on expensive, limited

Purpose of the Study:

• Introduce a de novo probabilistic metric for objective genome assembly quality assessment.
• Enable comparison of multiple assemblies from the same dataset without a reference genome.

Main Methods:

• Define assembly quality as the conditional probability of observing reads from the assembled sequence.
• Utilize a probabilistic score that is maximized by the true genome sequence.

Main Results:

• Demonstrate rapid and accurate computation of the de novo score, even for large datasets.
• Show that the de novo metric aligns with reference-based evaluations and outperforms traditional metrics like N50.
• Successfully applied the score to optimize genome assembler parameters, improving assembly quality.

Conclusions:

• Likelihood-based measures represent a new standard for de novo genome assembly evaluation.
• The proposed metric offers an objective and efficient approach to assessing genome assembly quality.