Efficient and accurate near telomere-to-telomere haplotype reconstruction of diploid genomes

Telomere-to-telomere (T2T) haplotype-resolved assembly represents the gold standard for capturing full diploid genetic diversity, enabling precise dissection of allele-specific expression, compound heterozygous disease mutations, and species evolutionary histories. However, recent landmark single-platform Oxford Nanopore (ONT) simplex read assemblers produce near-T2T assemblies plagued by pervasive haplotype chimerism across ambiguous genomic regions, rendering them unsuitable for long-range allele-specific analyses. This fundamental challenge arises from the inherent trade-off between two competing assembly paradigms: graph-based approaches excel at local haplotype resolution and ancestral segment delineation but lack chromosome-scale continuity, while sequence-based methods achieve long-range contiguity but fail to reliably separate phases across ambiguous genomic regions. Neither paradigm alone can simultaneously deliver both full-length contiguity and accurate phasing. To resolve this paradigmatic conflict, we present HapFold, the first hybrid assembly framework that synergistically integrates graph-based and sequence-based strengths with chromatin-contact data (Hi-C/Pore-C). Rather than merely concatenating these paradigms, HapFold unifies them into a single cohesive framework that overcomes the inherent limitations of each approach alone. HapFold processes unitig graphs from either PacBio HiFi or standard ONT simplex reads, delivering near-T2T haplotype assemblies with exceptional phasing accuracy. Comprehensive benchmarking across human, animal, and plant datasets demonstrates its superior performance over state-of-the-art methods. Notably, HapFold enables HiFi-derived assemblies to match the near-T2T contiguity of standard ONT-based assemblies without requiring additional ultra-long ONT sequencing. Additionally, it accelerates chromatin-contact mapping by an order of magnitude, significantly reducing computational costs and demonstrating robust performance with lower-coverage data. HapFold eliminates the long-standing trade-off between contiguity and phasing accuracy, providing a scalable solution for truly haplotype-resolved T2T assembly with broad applications in pangenomics, precision medicine, and agricultural breeding.