Informed kmer selection for de novo transcriptome assembly

Transcriptome assembly is one of the important step in many RNA-seq workflows. Currently de bruijn graph based de novo transcriptome assembly algorithms is widely used for this. One of the important parameter in any assembly algorithms is the k-mer (substrings of length k). The assembly generted by merging assemblies from different kmer values leads to optimal result when compared with assemblies generated by single kmer value. But there is no investigation on avoiding kmer values which doesnt really effect the final assembly. Currently researchers choose random kmer values spread over the read length. This leads to sub-optimal results and wastage of computational time and resources. We propose an algorithm named KREATION (Kmer Range EstimATION). The algorithms compares related assemblies and estimates the need of performing an additional kmer assembly. Based on a linear model fit, KREATION decides whether a kmer value beyond which no additional assembly is required. The algorithm was tested was datasets of differnet coverage and also with different assembly algorithms.  It is parameter free and also completely de novo.