COGRAM: A Computational Pipeline for Genome Assembly and Reconstruction via Optimized K-mer Sampling and De Bruijn Graph Networks

Genome assembly and annotation accuracy fundamentally depend on optimal selection of parameters and robust computational approaches. Here we introduce COGRAM (Coggins-Ramasamy Genomic Assembly Method), a novel bioinformatics pipeline that enhances genome assembly and reconstruction by optimizing k-mer parameters, leveraging graph theory, and incorporating machine learning techniques. Initially, COGRAM identifies the optimal k-mer length using methods inspired by KMERGENIE and grid search techniques, followed by random genomic sampling at the optimal resolution. It then conducts a comprehensive analysis of the frequency distributions of k-mer and GC-content across the sampled genome windows. Subsequently, the pipeline constructs a detailed de Bruijn framework graph from parsed genomic data. Using this graph, COGRAM trains a network to model genomic structures effectively, enhancing accuracy and scalability. Genome reconstruction is accomplished through rigorous cross-validation with a greedy algorithm designed to refine the quality of genome assembly iteratively. We demonstrate the effectiveness of COGRAM through benchmark tests on the E. coli genome. This pipeline represents a powerful tool for genomic projects with potential for expansion to other projects.