Prior knowledge on context-driven DNA fragmentation probabilities can improve de novo genome assembly algorithms

De novo genome assembly poses challenges when dealing with highly degraded DNA samples or ultrashort sequencing reads. Probabilistic approaches have been offered to enhance the algorithms, though existing methods rely solely on expected k-meric frequencies in the assemblies, neglecting the broader sequence context that strongly influences DNA fragmentation patterns.

Here, we present a proof of concept showing that prior knowledge on sequence context-driven DNA breakage propensities, through the dedicated parameterisation of k-mer assigned breakage probabilities, can be utilised to recover DNA assemblies that originate from fragmentation patterns more likely to have happened. Our approach is beneficial even for read lengths below the common ∼ 25 bp threshold of modern de novo genome assembly algorithms, and well below the threshold used for ultrashort fragments used in ancient DNA research.

This work could lay the groundwork for future enhanced de novo genome assembly algorithms, with improved ability to effectively assemble and evaluate ultrashort DNA fragments relevant for cell-free, ancient, and forensic DNA research.