Telomere-to-Telomere Phased Genome Assembly Using HERRO-Corrected Simplex Nanopore Reads

This project explores a strategy for generating telomere-to-telomere (T2T), haplotype-resolved genome assemblies using only ultra-long Oxford Nanopore (ONT) Simplex reads. We present HERRO, a deep learning-based error correction framework that significantly improves the accuracy of ONT Simplex reads while preserving haplotype and repeat distinctions. By applying HERRO-corrected reads in conjunction with the Verkko assembler, we produce highly contiguous assemblies, including full reconstruction of chromosomes X and Y in human diploid genomes. Our results demonstrate that high-quality phased assemblies can be achieved without the need for multiple long-read sequencing technologies. HERRO is compatible with both R9.4.1 and R10.4.1 ONT chemistries and generalizes well to other species.