CHM13-issues

CHM13 human reference genome issue tracking

For any downstream analysis, please use the following files:

• Possible consensus or mis-assembly issue: <ver.>_issues.bed
• Het sites: <ver.>/chm13.draft_<ver.>.curated_sv.20210612.vcf, <ver.>/chm13.draft_<ver.>.hets_combined.20210615.bed

Releases

• 2021-06-23 Updating 3 additional issues and adding error k-mers in v1.0 and v1.1
• 2021-06-15 Validated het SVs and clusters of heterozygous sites in v1.0 assembly
• 2021-04-28 Issues track for HiFi and ONT read alignments from Winnowmap 2.01
• 2021-03-08 Combined low coverage and clipped regions
• 2021-02-23 Low coverage regions for HiFi, CLR, and ONT read alignments

Alignments

HiFi and ONT reads were aligned to the assembly using Winnowmap 2.01. From the reads, top 0.02% of the repetitive 15-mers were collected using Meryl and downweighted in the process of minimizer sampling. Platform specific mapping parameters were provided; map-pb for hifi and map-ont for ONT reads, respectively. Reads were sorted and indexed using samtools, and filtered for primary reads with samtools view -F0x100.

Note: We did not perform this analysis on CLR reads as the information added was marginal or noisier compared to HiFi and ONT.

# Collect repetitive 15-mers
meryl count k=15 $ref output merylDB
meryl print greater-than distinct=0.9998 merylDB > repetitive_k15.txt

# Align
winnowmap --MD -W repetitive_k15.txt -ax map-$platform -t$cpus $ref $reads > $output.sam

# Sort and filter
samtools sort -@$cpus -m2G -O bam -o $output.bam $output.sam
samtools view -F0x104 -hb $output.srt.bam > $output.primary.bam

These alignments were converted to pairwise alignment format (paf) using paftools from minimap2 v2.17 and processed to collect low coverage and excessive clipped regions. Spurious alignment blocks shorter than 1kb or lower than 85% identity were removed from the ONT primary alignments.

# Convert to paf
samtools view -h -@$cpus $output.primary.bam | k8 paftools.js sam2paf - | cut -f 1-16 - >  $output.primary.paf

# Filter ONT for alignment blocks < 1kb and identity < 85%
awk '$11>1000 && $10/$11>0.85' ont.primary.paf > ont_pri.len1k_idy85.paf

Low coverage region

Supportive regions were collected for HiFi (>7x) and ONT (>10x) reads with asset commit v102, 0133f268eebf308a1c3eb356b564550526465157.

# Collect supportive regions from hifi
ast_pb -m 7 -M 10000000 -l 0 $paf > $pre.bed

# Collect supportive regions from ONT
ast_pb -m 10 -M 10000000 $paf > $pre.bed

Alignment blocks were trimmed 300bp on both sides to avoid spurious mapping while collecting coverage in ONT. Low coverage regions were obtained with bedtools subtract, and merged when regions were 5kb apart with bedtools merge -d 5000. Regions overlapping collapsed rDNA region (only applied in v1.0 assembly) and 3kb around the end of chromosomes were excluded.

To distinguish the cause of low coverage, we collected % GA/TC, GC, and AT micro satellite repeats in every 128 bp window in the assembly that appears as dimers when compressing homopolymers. Any low coverage region overlapping with >80% of any micro satellite repeat within 10kb distance was marked accordingly.

In addition, regions overlapping with known consensus issues obtained with Merqury Illumina-HiFi hybrid 21-mers were marked as Low_Qual.

Clipped region

Total number of reads with more than 100 bp soft-clipped or hard-clipped bases were collected in every 1024 bp window for both HiFi and ONT reads. Clipped regions were identified when 1) >10x HiFi or ONT reads were clipped; or 2) >10% of the HiFi or >15% of the ONT total aligned reads were clipped.

Associated files

• <ver.>_issues.bed: Low coverage not associated with sequencing biases and other issues

Label	Description	R,G,B	Color
Low	Low coverage	204,0,0	red
Low_Qual	Low coverage from lower consensus quality	204,0,0	red
Error_Kmer	K-mers identified as errors from the Illumina-HiFi hybrid 21-mers	0,0,0	black
Collapse	Approximate region conatining sequence collapse	204,0,0	red
Chimeric_Hap	Chimeric consensus of two haplotypes	204,0,0	red

• issues_raw/<platform>.issues.bed : Low coverage + Flanking clipped regions

Label	Description	R,G,B	Color
Low	Low coverage	204,0,0	red
Low_Qual	Low coverage from lower consensus quality	204,0,0	red
Low_GA/TC, Low_AT	Low coverage from sequencing bias	153,153,255	light purple
Clipped	Region with excessive read clipping	153,153,153	gray

Other files under issues_raw:

• clipped.bed
• Absolute num. reads with clipping: hifi_pri.w1k.clip_abs.wig, ont_pri.len1k_idy85.w1k.clip_abs.wig
• Relative fraction of clipped reads compared to all reads: hifi_pri.w1k.clip_norm.wig, ont_pri.len1k_idy85.w1k.clip_norm.wig
• Value of -1 represents windows with no read alignments

Het sites

Heterozygous sturcutral variants and clusters of heterozygous regions collected on v1.0 release.

A manuscript describing the details of each variant calling methods are in preparation and will be posted soon.

Clusters of heterozygous regions: chm13.draft_<ver.>.hets_combined.YYYYMMDD.bed

Abbreviation	Description
ID	Numbered ID
SC	Source of detection method
AF	Allele frequency
PL	Platform support
LEN	SV size
(TYPE)	INS, DEL, DUP, INV for Insertion, Deletion, Duplication, Inversion

SC	Source
Curated	Hets from SV calling methods, manually curated for the position, LEN, AF, and SV type (manusript in prep.)
Clipped	Hets from read clippings in HiFi and ONT, curated for position, LEN, AF, and SV type (manusript in prep.)
NucFreq	Clusters of heterozygous small variants called with NucFreq
tQ	Heterozygous SVs and AF detected from TandemMapper and TandemQUAST

Curated SVs: chm13.draft_<ver.>.curated_sv.YYYYMMDD.vcf

SVs manually validated for breakpoints, CCS AF, and SV type. CLR and ONT AF may be slightly lower than the actual AF due to mapping biases.

Citation

Please cite the paper below if any of the materials posted on this github are used:

Nurk, Koren, Rhie, and Rautiainen et al, The complete sequence of a human genome, bioRxiv (2021) https://doi.org/10.1101/2021.05.26.445798