Routine automated gapless haplotype-resolved assembly of any human genome, including rDNA arrays
A complete diploid human genome now exists and HPRC Release 2 delivers 232 assemblies, so what remains is automation, rDNA arrays and curation: engineering-heavy and the easiest problem in the area (39), comparable to the verified-multiplier problem (41). Application value 40 (rare disease and pangenomics, with short-read pipelines already serving most clinics); leverage 38 as reference quality underpins variant calling.
The problem
Reading a person's entire genome, both parental copies, from end to end with no gaps, automatically.
- 1Long reads from sequencer
- 2Overlaps become a graph
- 3Parental copies separated
- 4Long repeats must be spanned
- 5Gapless, curated by software
State of play. T2T-CHM13 (2022) completed a haploid human genome. The Human Pangenome Reference Consortium's first release had 47 phased diploid assemblies; Release 2 (May 2025, pre-publication) has 232 individuals made from PacBio HiFi and Oxford Nanopore ultra-long reads, phased with parental or Hi-C data, with machine-learning polishing and errors below about one per 500 kb. A bioRxiv benchmark (T2T-HG002, 2025; reported in Cell 2026) is telomere-to-telomere for all 46 chromosomes except inside the rDNA arrays, with 99.4 percent of the genome free of detectable errors.
Path to understanding
foundations → problemThe research entries this problem rests on, in the order to read them. Each links to its explainer.
- Start with
- Step 2
- Fast matrix multiplicationAlgorithm
- Step 3
- Step 4
- Genome sequence alignment and assemblyAlgorithm · host
- Then
- Routine automated gapless haplotype-resolved assembly of any human genome, including rDNA arraysThe problem
State of the art
- T2T-CHM13 (Science 2022). The first gapless human genome, from a haploid hydatidiform mole, finished with extensive manual effort and multiple technologies.
- HPRC Release 1 (Nature 2023). 47 phased diploid assemblies covering more than 99% of the expected sequence at more than 99% accuracy.
- HPRC Release 2 (announced 12 May 2025, pre-publication). 232 individuals from HiFi and ultra-long nanopore data with trio or Hi-C phasing and DeepConsensus plus deep-learning polishing; misassemblies reduced about 3-fold, single-base and short-indel errors about 2-fold (below one base error per 500 kb), and a substantial number of complete T2T chromosomes, but the release is not entirely gapless and the exact complete-chromosome fraction was not stated on the release page reviewed.
- T2T-HG002 (bioRxiv preprint, 2025; a journal version is reported but was not verified for this entry). A complete diploid benchmark with T2T continuity for all 46 chromosomes except the interior of the rDNA arrays, with 99.4% of the diploid genome free of detectable errors and 701.4 Mb autosomal sequence plus 216.8 Mb on sex chromosomes absent from previous benchmarks. It required extensive curation; the abstract notes state-of-the-art de novo assembly resolves 2-7% more sequence than earlier approaches and has about one error per 100 kb in 99.9% of benchmark regions.
- Gap. Fully automated, curation-free T2T assembly of arbitrary individuals, with validated rDNA arrays, has not been demonstrated at population scale.
Realistic applications
Complete personal diploid genomes would reveal structural variants, repeat expansions and haplotype-specific regulation missed by short-read pipelines, improve diagnosis of rare disease, enable population-scale pangenome references that reduce reference bias, and make rDNA and satellite variation a measurable trait.
What counts as solved
A refereed demonstration meeting criteria (i)-(v) across at least 100 individuals, using released assembler software and standard sequencing data, with independent validation of rDNA copy number and with the assemblies deposited in public archives.
How it gets verified
ReproductionA training run, experiment or computation rerun by someone else.
Released assembler software on standard sequencing data must meet criteria (i) to (v) across at least 100 individuals, with independent validation of rDNA copy number and the assemblies deposited in public archives.
Claims under verification
+ add a claimNo claimed results on the Verification board for this problem yet.
Where the work is
Mostly sequencing and assembler engineering plus validation of the hardest repeats.
Bottlenecks
ranked by votes- #1EngineeringrDNA and satellite arrays
Megabase-scale near-identical repeats remain unresolved in T2T-HG002 (rDNA interiors) and limit fully automated assembly.
Community grades
Sign in to add your own grade.
Connected problems
Progress
- Sep 21, 2025T2T-HG002 diploid benchmarksource
All 46 chromosomes T2T except rDNA array interiors; 99.4% error-free by benchmark metrics.
- May 12, 2025HPRC Release 2source
232 individuals, with large gains in continuity and accuracy.
- May 10, 2023HPRC draft pangenomesource
47 phased diploid assemblies at over 99% accuracy.
- Mar 31, 2022T2T-CHM13source
First complete human genome sequence (haploid).