1. Karen Miga presented on reaching complete telomere-to-telomere chromosome assemblies using long-read sequencing. 2. The current human reference genome is incomplete, with gaps and unresolved issues remaining. 3. Miga's lab generated a high-quality assembly of chromosome 21 for the CHM13 genome using Oxford Nanopore long reads totaling 155 GB of sequence data.

1. Telomere-to-telomere assembly of a
complete human chromosomes
Karen Miga
UC Davis Genetics Seminar
Sept 30, 2019
@khmiga

2. New Era in Genetics and Genomics
We are finally reaching complete, high-quality
telomere-to-telomere chromosome assemblies

3. New Era in Genetics and Genomics
We are finally reaching complete, high-quality
telomere-to-telomere chromosome assemblies
Human reference genome is incomplete.
• 368 unresolved issues, 102 gaps
• Segmental duplications, gene families, satellite
arrays, centromeres, rDNAs
• Uncharacterized sequence variation in the human
population

4. New Era in Genetics and Genomics
We are finally reaching complete, high-quality
telomere-to-telomere chromosome assemblies
Human reference genome is incomplete.
• 368 unresolved issues, 102 gaps
• Segmental duplications, gene families, satellite
arrays, centromeres, rDNAs
• Uncharacterized sequence variation in the human
population
chr21

5. New Era in Genetics and Genomics
We are finally reaching complete, high-quality
telomere-to-telomere chromosome assemblies
Human reference genome is incomplete.
• 368 unresolved issues, 102 gaps
• Segmental duplications, gene families, satellite
arrays, centromeres, rDNAs
• Uncharacterized sequence variation in the human
population
Our current understanding of
genome biology and function30 Mb
chr21

6. New Era in Genetics and Genomics
We are finally reaching complete, high-quality
telomere-to-telomere chromosome assemblies
Human reference genome is incomplete.
• 368 unresolved issues, 102 gaps
• Segmental duplications, gene families, satellite
arrays, centromeres, rDNAs
• Uncharacterized sequence variation in the human
population
Our current understanding of
genome biology and function30 Mb
chr21
~20 Mb ?

7. Challenge:
Generating assemblies across repetitive regions that
span hundreds of kilobases.
Repeats (100 kb+)
Unique
variant
Unique
variant
Can high-coverage ultra-long sequencing resolve
complete assemblies of the human genome?

8. MinION
100kb+

9. It’s time to finish the human genome
The Telomere-to-Telomere (T2T) consortium is an
open, community-based effort to generate the
first complete assembly of a human genome.

10. Our target: CHM13hTERT
Cell line from Urvashi Surti, Pitt; SKY karyotype from Jennifer Gerton and Tamara Potapova, Stowers
N=46; XX

11. Our target: CHM13hTERT
Cell line from Urvashi Surti, Pitt; SKY karyotype from Jennifer Gerton and Tamara Potapova, Stowers
N=46; XX

12. Intramural Sequencing Center
CHM13 Sequencing
94 MinION/GridION flow cells
11.1M reads
155 Gb (1.6 Gb / flow cell) (50x)
99 Gb in reads >50 kb (32x)
78 Gb in reads >70 kb (25x)
Max mapped read length 1.04 Mb
From May 1/18 – Jan 8/19

13. Intramural Sequencing Center
CHM13 Sequencing
94 MinION/GridION flow cells
11.1M reads
155 Gb (1.6 Gb / flow cell) (50x)
99 Gb in reads >50 kb (32x)
78 Gb in reads >70 kb (25x)
Max mapped read length 1.04 Mb
From May 1/18 – Jan 8/19
50x Nanopore ultra-long
Contig building
60x PacBio
Polishing
50x 10x Genomics
Polishing
BioNano
Structural validation

14. • 2.94 Gbp assembly NG50: 75 Mbp
• Exceeds the continuity of the reference
genome GRCh38 (56 Mbp NG50
contig size).
• Subset of chromosome assemblies
break only at centromere.
Roadmap for completing the genome
Canu

15. Canu

16. Canu

17. Orthogonal Validation
Jo and Valerie

19. 2.2 - 3.7 Mb
mean of 3010 kb (S.D. = 429; n = 49)

20. STRUCTURAL VARIANT

21. STRUCTURAL VARIANT
151516 15 3 8 2
8
4
Assemble contigs
Using overlapping
SV patterns

22. XqXp
Scaffold Assembly of XCEN

23. XqXp
Rel3 Assembly: ~3.1 Mb
The assembly is a hypothesis(!)

24. 2107 294659
Beth SullivanJennifer Gerton
Edmund Howe
Rel3 Assembly: ~3.1 Mb

25. @NanoporeConf | #NanoporeConf
Marker-assisted mapping
Adam Phillippy Arang Rhie Sergey Koren

26. @NanoporeConf | #NanoporeConf
Create a scaffold of unique, or
single copy k-mers genome-wide
Marker-assisted mapping
Adam Phillippy Arang Rhie Sergey Koren
Marker-assisted mapping

27. @NanoporeConf | #NanoporeConf
Anchor high-confident
long-read alignments to
repeat assemblies
Marker-assisted mapping
Adam Phillippy Arang Rhie Sergey Koren
Marker-assisted mapping

28. 28
Confident mapping of long reads
using a single-copy k-mer strategy
Identify and mark all sites of unique anchors across the chromosome
chrX
• 21-mers that appear ~c times in Illumina data
• Also found in PacBio/Nanopore reads
• Less frequent in the centromere, but still there
• (Validated with Duplex-Seq)

29. 29
Confident mapping of long reads
using a single-copy k-mer strategy
Filter long read alignments: retaining those with unique k-mer anchoring
chrX
chrX

30. 30
Spacing of single-copy k-mers can be irregular in
repeat-dense regions
chrX
chrX
X CENTROMERE ARRAY
CENTROMERE
CENX: 3.1 Mbps
Number of k-mers: 2,034
Spacing N50: 6,879
Longest distance
between k-mers
: 53,798 bp

31. 31
10XG Polishing
Unique K-mer-based filtering: Nanopore Reads
longranger + freebayes (two rounds)
nanopolish (two rounds)
arrow (two rounds)
Unique K-mer-based filtering: PacBio (CLR) Reads
chrX
chrX
chrX

32. GAGE pre-polishing
ChrX GAGE array: 19 tandemly arrayed ~9.4 kb repeats
Coverage
250
200
150
100
50
0
Base position
Most frequent base
Second most frequent base (error)
19 tandemly arrayed ~9.4 kb repeats

33. GAGE with marker-assisted polishing
Most frequent base
Second most frequent base (error)
ChrX GAGE array: 19 tandemly arrayed ~9.4 kb repeats
Coverage
250
200
150
100
50
0
Base position
19 tandemly arrayed ~9.4 kb repeats

34. 34
CSS/HiFi Evaluation
chrX
HiFi Alignments to Evaluate Polishing
CENTROMERE X:
BEFORE POLISHING
DXZ1: 3.1 Mb

35. 35
CSS/HiFi Evaluation
chrX
HiFi Alignments to Evaluate Polishing
CENTROMERE X:
AFTER POLISHING
NOTE:
Underlying satellite array
structure remains the same.
DXZ1: 3.1 Mb

36. Opens the whole genome to analysis
Ariel Gershman
Winston Timp’s
Laboratory

37. Ariel Gershman
Winston Timp’s
Laboratory

38. Ariel Gershman
Winston Timp’s
Laboratory

39. Ariel Gershman
Winston Timp’s
Laboratory

40. 1. Structurally validated assembly from telomere-to-telomere. Including
3.1 Mb tandem repeat at the X centromere and providing a complete
assessment across tandemly repeated gene families.
Finished T2T X Chromosome:
High Accuracy and High Continuity

41. 1. Structurally validated assembly from telomere-to-telomere. Including
3.1 Mb tandem repeat at the X centromere and providing a complete
assessment across tandemly repeated gene families.
2. Novel polishing strategy capable of improving the quality of large repeat-
rich regions. Demonstrating dramatic improvements in quality over the
entirety of the X chromosome.
Finished T2T X Chromosome:
High Accuracy and High Continuity

42. 1. Structurally validated assembly from telomere-to-telomere. Including
3.1 Mb tandem repeat at the X centromere and providing a complete
assessment across tandemly repeated gene families.
2. Novel polishing strategy capable of improving the quality of large repeat-
rich regions. Demonstrating dramatic improvements in quality over the
entirety of the X chromosome.
3. Statistics of CHM13 full length BAC alignments to polished assembly:
275/341 (81%) QV 37.4 QV 27.9
153/341 (45%) QV 37.7 QV 27.4
Vollger M, Logsdon, G et al. bioRxiv doi.org/10.1101/635037
MeanMedianBACs Aligned
HiFi
UL-asm
Finished T2T X Chromosome:
High Accuracy and High Continuity

43. @NanoporeConf | #NanoporeConf
It is time to finish the
human genome

44. • github.com/nanopore-wgs-consortium/chm13
• 120x Nanopore reads
• NHGRI, UW, Nottingham,
• UC Davis (PromethION, Megan Dennis)
• 50x 10x Genomics linked reads (NHGRI)
• 70x PacBio CLR reads (WashU)
• 24x PacBio HiFi reads (UW)
• 40x Hi-C (Arima Genomics)
• BioNano optical map (WashU)
• Unpolished Canu assemblies
NEW! Rel3 open data release

45. Additional ultra-long ONT data
from Glennis Logsdon (UW)
Read length Coverage Percent of data
>50 kbp 12X 86%
>100 kbp 9.1X 66%
>150 kbp 6.8X 49%
>200 kbp 4.9X 35%
>250 kbp 3.4X 24%
N50 = 147.1
N1 = 649.6
Max = 1538.3
0.1 1 10 100 1000 10,000
Read length (kbp)
20,000
17,500
15,000
12,500
10,000
7,500
5,000
2,500
0
Numberofreads
13.9X coverage
• github.com/nanopore-wgs-consortium/chm13

46. • Minimal change in continuity
• 79.5 Mbp (rel2) vs. 71.8 Mbp (rel3) NG50
• Don’t judge assemblies based on continuity
• Tricky regions are fixed
• GAGE and more SegDups automatically resolved
• Improved BAC validation
• 288 (rel2) vs. 310 (rel3) of 341 BACs resolved
• 1 chromosome down, 23 to go…
Triple the coverage, what changed?

47. Goal of a complete human genome in the next two
years.
Challenges in front of us:
• Acrocentric p-arms
• Large segmental duplications
• Classical Human satellites 2,3
Establishing new benchmarking standards (XChr)
Pioneering new pipelines: Polishing, repeat assembly, and array
structural validation.
Setting the bar higher for quality and completeness.