Telomeres have variable length across chromosomes, and across cells. The reads generated via whole genome sequencing (WGS) can be processed to compute the avarage length of telomeres across chromosomes and within a cell population. Computel uses the following steps to estimate MTL:

Step 1. Distinguish telomeric reads from non-telomeric ones

Telomeric reads are those reads that are sequenced from the telomeric regions of the chromatin or from junction of telomeres and subtelomeric regions.

In order to distinguish between the reads Computel first performs step 0.

Step 0. Alignment of reads to a telomeric reference

Unlike equivalent methods for identification of telomeric reads that are based on pattern matching, Computel uses alignment to a specially designed telomeric reference, using Botwie 2. Due to alignment based identification, Computel is on one hand robust against sequencing errors, and sequence variations, and on the other hand it minimizes the number of false positive hits that could come from non-telomeric reads containing telomeric repeats (interstitial telomeric sequences).

Step 2. Compare the number of telomeric reads with sequencing coverage to compute MTL

The ratio of the coverage at telomeric reference  and that at the genomic reference is proportional to the number of telomeric ends and the telomere length and is inversely proportional to read length.

MTL = rel.cov * (rl+pl-1) / (2*n_chr), where: rel.cov  = tel.cov / base.cov;   tel.cov – mean telomere index coverage;   base.cov – mean reference genome coverage; rl  – read length; pl – telomeric pattern length; n_chr – number of chromosomes in a haploid genome

Telomeres do not always possess canonical repeat patterns. For example, in humans, the canonical repeat is ‘TTAGGG’. The telomeres may also contain variants of this pattern, with insertion, deletions and substitutions. The relative abundance and composition of telomeric variants may change during disease development or cellular proliferation. It is thus important to be aware of the distribution of telomeric repeat variants in the studied samples, along with computing their MTL.

Computel uses the telomeric alignment files as input, to estimate abundance of canonical and variant telomeric repeat patterns. The CIGAR strings in the sam files guide Computel for a proper frame to identify repeat variants, and the quality values of the mismatched or inserted bases indicate whether the variant has likely arisen from sequencing errors or from mutations.

System requirements

You have to have R (version 3.0.3 or higher) and samtools version 1.3 or higher (for Computel v0.4) installed in your system.

Installation

For installation download and uncompress the Computel package in a local directory. The required binaries and files for setup configuration are set in the package. Make computel.sh executable by running ‘chmod +x computel.sh’.

Testing

To test if Computel is set up properly, we advise to run the following command first:

computel.sh -1 src/examples/tel_reads1.fq.gz -2 src/examples/tel_reads2.fq.gz -o mytest

Usage

The basic usage is:

./computel.sh [options] -1 <fq1> -2 <fq2> -o <outputpath>

<fq1> fastq file (the first pair or the only fastq file (for single end reads)

<fq2> fastq file (optional: the second pair of fastq files, if exists)

<fq3> fastq file (optional: the third pair of fastq files, if exists)

<o> output directory (optional: the default is computel_out)

Type ./computel.sh in the terminal for additional options

Output

Two files are generated after successful run:

1. tel.length.xls – this file contains information about the provided reads (length, base coverage), organism info (telomere pattern, genome length, number of haploid chromosomes) and the telomere length computed by Computel.


2. tel.variants.xls – contains the absolute and relative abundance of the number of telomeric repeat patterns

More options

More details and additional options are described at the Computel github repository.

Currently available tools for MTL measurement from WGS do pattern matching to distinguish between telomeric and non-telomeric reads. Computel uses alignment approach. We align the reads to a specially designed telomeric reference. This makes Computel’s algorithm more robust against sequencing errors and telomeric variants arising in different pathological conditions.

Computel outputs MTL in absalute values, in base pair units.

Yes. The default species for Computel is homo sapiens, with default species-specific parameters: telomeric repeat pattern: “TTAGGG” number of haploid chromosomes: 23 genome length: 3244610000 If you’d like to suppl another organism’s WGS data, you should change these parameters with Computel’s options respectively.

Due to the flanking N’s at only one side of the telomeric reference, Computel is able to minimize the number of interstitial telomeric sequences, wrongly classified as telomeric. This makes Computel more specific than other pattern-matchin algorithms.

Computel is designed to increase its sensitivity by reducing the amount of false negative hits: i.e.  unidentified telomeric reads. Compared to pattern-matching algorithms, mismatch mutations and small INDELs do not lead to false negative results, as those still allow for allingment to the telomeric reference. The alignment by bowtie 2 may itself lead to low sensitivity, thus, we have carefully calibrated the alignment options to increase sensitivity (at the same time not affecting specificity).

Computel increases its specificity by reducing the number of genomic reads having telomeric repeats (a.k.a. interstitial telomeric sequences). This is achieved  due to the flanking N’s at only one side of the telomeric reference. This makes Computel more specific than other pattern-matchin algorithms.

For tests giving exact values of sensitivity and specificity depending on read length and genomic coverage, please, refer to the supporting information of our paper.

No. In comparison with pattern-matching algorithms, Computel checks for any variants present in the telomeric reads. Due to using CIGAR strings generated after the alignment, Computel reduces variant redundancy that could otherwise arise due to frameshift pattern matches (this will be documented in Computel version 2.0 manuscript).

Computel checks that identified variants pass a certain quality check, based on read qualities provided in the FASTQ files. We have chosen a quality threshold that leads to most accurate results (this will be documented in Computel version 2.0 manuscript).

Min.seed is the seed parameter given to bowtie 2 for alignment. The more the min.seed, the more the specificity and less the sensitivity of Computel. We have performed multiple tests on reads ranging between 20-150 bp in length, and found that min.seed of default value 12 is ideal. We don’t recommend changing this value, unless you have a clear reason for doing so.

Computel does not explicitly account for aneuploidy or any other changes in the number of telomeric ends or extra-chromosomal telomeric sequences. However, if you are aware of the changes in the number of chromosomes in your samples, you can change the number of haploid chromosomes in the Computel’s input and this will let you adjust the mean telomere length value you obtain.