PankratzLab / NGS-TL

Estimate telomere length (TL) from whole genome sequencing (WGS)

NGS-TL

Estimate telomere length (TL) from whole genome sequencing (WGS)

Example usage

The primary entry point is to run ngsTL.sh contained in the docker image (https://github.com/PankratzLab/NGS-TL/pkgs/container/ngs-tl)

Using singularity

singularity run \
...
"docker://ghcr.io/pankratzlab/ngs-tl:main" \
/app/NGS-TL/ngsTL.sh \
# The cram (or bam) file to process
--cramFile "$cramFile" \
# The cram (or bam) file's index
--craiFile "$craiFile" \
# The root output path (e.g. rootOutput=/path/to/example should result in the creation of /path/to/example.ltl.estimate.txt.gz)
--rootOutput "$rootOutput" \
# The reference genome used to align the cram or bam (e.g. GRCh38_full_analysis_set_plus_decoy_hla.fa)
--referenceGenome "$referenceGenome" \
# Stores 1kb bins that have gc content similar to telomeric repeats.
# Unless you want to change it, this can be set to https://github.com/PankratzLab/NGS-TL/blob/main/resources/GRCh38_full_analysis_set_plus_decoy_hla.1kb.LTL.GC.filtered.bed.gz
--gcBedFile "$gcBedFile" \
# A bed file defining the regions to look for telomeric reads (typically 25kb from the ends of each chromosome)
# Unless you want to change it, this can be set to https://github.com/PankratzLab/NGS-TL/blob/main/resources/25kb.bins.bed
--regionsSearch "$regionsSearch"



# (Optional) If a mosdepth file is provided, mosdepth will not be run.
# If mosdepth file is not provided, mosdepth will be run to produce $rootOutput.regions.bed.gz
#--mosdepthFile $pathToMosdepthFil

Using dsub

The command-line tool dsub can be run using arguments and invocation as the above singularity example (the below is untested, and not succesfully run):

dsub ... --image "ghcr.io/pankratzlab/ngs-tl:main" ...

Example script

A parameterized example with all expected output can be found in the example directory of this repo. Output was generated using example.sh

Output

The primary output of interest is a file named "$rootOutput".ltl.estimate.txt.gz (e.g. https://github.com/PankratzLab/NGS-TL/blob/main/example/NA12878.ltl.estimate.txt.gz).

This is an example from NA12878.ltl.estimate.txt.gz:

mosdepth_gc_coverage_min	mosdepth_gc_coverage_max	mosdepth_gc_coverage_mean	mosdepth_gc_coverage_sd	mosdepth_gc_coverage_median	mosdepth_gc_coverage_mad	mosdepth_gc_coverage_iqr	mosdepth_gc_coverage_n	READ_LENGTH	TL_READS_AT_K_1	LENGTH_ESTIMATE_AT_K_1	TL_READS_AT_K_2	LENGTH_ESTIMATE_AT_K_2	TL_READS_AT_K_3	LENGTH_ESTIMATE_AT_K_3	TL_READS_AT_K_4	LENGTH_ESTIMATE_AT_K_4	TL_READS_AT_K_5	LENGTH_ESTIMATE_AT_K_5	TL_READS_AT_K_6	LENGTH_ESTIMATE_AT_K_6	TL_READS_AT_K_7	LENGTH_ESTIMATE_AT_K_7	TL_READS_AT_K_8	LENGTH_ESTIMATE_AT_K_8	TL_READS_AT_K_9	LENGTH_ESTIMATE_AT_K_9	TL_READS_AT_K_10	LENGTH_ESTIMATE_AT_K_10	TL_READS_AT_K_11	LENGTH_ESTIMATE_AT_K_11	TL_READS_AT_K_12	LENGTH_ESTIMATE_AT_K_12	TL_READS_AT_K_13	LENGTH_ESTIMATE_AT_K_13	TL_READS_AT_K_14	LENGTH_ESTIMATE_AT_K_14	TL_READS_AT_K_15	LENGTH_ESTIMATE_AT_K_15	TL_READS_AT_K_16	LENGTH_ESTIMATE_AT_K_16
0	126.56	4.835327	2.169083	4.87	1.082298	1.47	24563	101	16421	101.00336363383	15285	94.0159803387793	15140	93.1241048301681	15014	92.3490957675128	14846	91.315750350639	14657	90.1532367566561	14419	88.6893307494183	14175	87.1885195487207	13880	85.3740142036151	13522	83.1720043271818	13094	80.5394338603844	12610	77.5624149212959	11986	73.7242748014791	11280	69.3817637043788	10227	62.9049022521882	8087	49.7420499182014

Brief data dictionary of the results:

Column	Description
mosdepth_gc_coverage_min	The minimum coverage of the 1kb mosdepth bins that overlap the gcBedFile
mosdepth_gc_coverage_max	The max coverage of the 1kb mosdepth bins that overlap the gcBedFile
mosdepth_gc_coverage_mean	The mean coverage of the 1kb mosdepth bins that overlap the gcBedFile. Currently, this value is used for normalization
mosdepth_gc_coverage_sd	The sd of the coverage of the 1kb mosdepth bins that overlap the gcBedFile
mosdepth_gc_coverage_median	The median coverage of the 1kb mosdepth bins that overlap the gcBedFile
mosdepth_gc_coverage_mad	The mad of the coverage of the 1kb mosdepth bins that overlap the gcBedFile
mosdepth_gc_coverage_iqr	The iqr of the coverage of the 1kb mosdepth bins that overlap the gcBedFile
mosdepth_gc_coverage_n	The number of 1kb mosdepth bins that overlap the gcBedFile
READ_LENGTH	Empircially determined read length of the bam (max read length reported by samtools stats)
TL_READS_AT_K_1	Number of reads containing at least 1 telomeric repeat
LENGTH_ESTIMATE_AT_K_1	TL estimate using a repeat "K" threshold of 1
TL_READS_AT_K_2	Number of reads containing at least 2 telomeric repeats
LENGTH_ESTIMATE_AT_K_2	TL estimate using a repeat "K" threshold of 2
…	…
TL_READS_AT_K_N	Number of reads containing at least N telomeric repeats
LENGTH_ESTIMATE_AT_K_N	TL estimate using a repeat "K" threshold of N

Which LENGTH_ESTIMATE_AT_K to use?

TODO ... but the individual length estimates can be evaluated against a PRS and the estimate with the strongest association could be used as aproxy.

Brief methods

The general strategy is similar to TelSeq, with a few shortcuts for speed (TelSeq reference https://www.ncbi.nlm.nih.gov/pmc/articles/PMC4027178/)

Here is a brief overview of the methods currently utilized in this repository to estimate TL from WGS. ngsTL.sh runs the following steps.

1. mosdepth.sh : If a mosdepth file is not provided up front, generate coverage of each 1kb bin across the entire genome via mosdepth
2. extractMosdepthGC.sh: Extract regions with gc content similar to telomeres (48% <-> 52%) from the mosdepth 1kb bins. Limit these bins to those that do not overlap the set of lower quality bins from NGS-PCA (see https://github.com/PankratzLab/NGS-PCA#exclude-bed for more details). Then compute mean coverage of those regions.
3. extractEnds.sh: Extract reads from the cram that have been mapped to within 25kb of chromosomal ends (regions defined by 25kb.bins.bed)
4. countTLReads.sh: Count the number of TTAGGG/CCCTAA telomeric repeats seen in each read extracted from the previous step. Instead of scanning the entire cram/bam as in TelSeq, we only look for telomeric reads near chromsome ends
5. estimateTL.sh: Estimate TL using mean coverage of GC bins from step 2 and the number of TL reads from step 4. Using estimate.R, a TL estimate is computed for each TL "K" (TL_READS_AT_K in the output above)