SexChrLab / SexInference

Infer sex from DNA sequencing data and RNA sequencing data

SexInference

Infer sex from DNA sequencing data and RNA sequencing data

DNAseq

• Scripts are under DNAseq
• General idea:
  • Empirically, we observe that when mapping a genetic XX female individual to a reference genome, the ratio in reads mapped between the X chromosome and a chromosome on the autosomes (for example, chr19) is relatively high (>50%) while the ratio in reads mapped between the Y chromosome and chr19 and the ratio in reads mapped between the Y chromosome and the X chromosome is low (<1%).
  • On the other hand, we observe that when mapping a genetic XY female individual to a reference genome, the ratio in reads mapped between the X chromosome and chromosome 19 is low (<50%) while the ratio in reads mapped between the Y chromosome and chr19 and the ratio in reads mapped between the Y chromosome and the X chromosome is relatively high (at least greater than 5%)
  • To infer sex from DNAseq data, we are going to leverage these observation and make prediction about whether the sample is likely to be XX female or XY male bsaed on the ratio in reads mapped between the X chromosome and chr19, between the Y chromosome and chr19, and between the X and the Y chromosomes.

Required programs:

• Samtools 1.7
• Python 3.8.5

Step 1: Download the github repo:

git clone https://github.com/SexChrLab/SexInference.git

Step 2: Infer sex from mapped bam files

• The script takes in as input a bam file that has been mapped to a default reference genome. If you start with the fastq files, you need to map the fastq files to an appropriate reference genome.

• This is how the script is run:

python infer_sex_from_readsmapped.py --sample {sample_id} --bam_path {/path/to/bam/} --out_dir {/path/to/output/directory}

• After the script is successfully run, it produces a directory with the same name as the sample id inside the output directory. 4 files are found under this directory:
  • chr19_bam_stat.txt: output from running samtools stats on chr19
  • chrX_bam_stat.txt: output from running samtools stats on chrX
  • chrY_bam_stat.txt: output from running samtools stats on chrY
  • {sample_id}_summary.tsv: a summary file with the following columns:
    • sample: sample id
    • chr19: number of reads mapped for chr19
    • chrX: number of reads mapped for chrX
    • chrY: number of reads mapped for chrY
    • X/19: ratio in reads mapped between chrX and chr19
    • Y/19: ratio in reads mapped between chrY and chr19
    • Y/X: ratio in reads mapped between chrY and chrX
    • sex_prediction: either probable_XX or probable_XY

How to use this script:

• We anticipate that there are 2 ways that you can use this script for.
• If you want to know the probable sex of a sample, you can run the script for your sample, and check the summary file to see what we think the probable sex to be, based on our observation of X/19, Y/19, and Y/X ratios
• If you have a lot of samples in your study, and you have the information whether the sample was reported as male or female, you can run the scripts for all of these samples and plot X/19, Y/19 and Y/X ratios for all of the reported female samples and for all of the reported male samples. This visualization allows you to check if there's any sample that are different from the rest of the samples in terms of X/19, Y/19, and Y/X ratios.

RNAseq

• Scripts are under RNAseq

Required programs:

• These R packages are needed to run: tximport, GenomicFeatures, tidyr, tidyverse, dplyr, caret, and glmnet.

In Human

• General idea: we observed that gene expressions from XIST (female-specific) and Y-linked genes IF1AY, KDM5D, UTY, DDX3Y, and RPS4Y1 are indicative of sex. Therefore, to infer sex from RNAseq data, we fit a logistic regression model using gene expression from these six genes as predictor variables using the GTEx data and predict sex on any new data based on the model build on the GTEx data.
• Below are the steps to infer sex on RNAseq data for any new dataset:

Step 1: Download the github repo

• The github repo contains the training data required for the model

git clone https://github.com/SexChrLab/SexInference.git

Step 2: Run quantification using salmon

• You will need to run salmon on your samples but see script salmon_snakemake.snakefile for an example of how to set up salmon.
• The pipeline currently works with the reference transcriptome gencode.v29.transcripts.fa (https://www.gencodegenes.org/human/release_29.html). It is important that this version of the reference genome is used when running quantification because we are matching the transcript IDs to the gene IDs for this version of the transcriptome in Step 3. If you are using a different version of the transcriptome, please see where to edit the gene names in Step 3 below.

Step 3: Convert transcript level to gene level

• Because salmon generates counts for transcripts, and because we find that expression at the gene level is clearer for predicting sex, we convert transcript level to gene level for the 6 genes using the script tximport.R. This script outputs a CSV file with 7 columns. The first 6 columns are the counts for the genes XIST, IF1AY, KDM5D, UTY, DDX3Y, and RPS4Y1 and the last column is the reported sex.

• Usage:

  Rscript tximport.R {path/to/gtf/file} {path/to/samples/name} {quant/dir} {outfile}
  

  • Argument 1: path to the gencode.v29.annotation.gtf or the annotation that matches with the reference genome to use in Step 1. The matching ID between the transcript and the gene is currently based on gencode v29. Notes: if you are using a different transcriptome in Step 2, you will need to check if the transcript ID match with the gene ID for these 6 genes. Check lines 38-43 of the script tximport.R and modify the transcript ID and gene ID as necessary.
  • Argument 2: path to the samples name. The first column is the sample name and the second column is the reported sex. See an example under the folder example_data (samples_name.txt).
  • Argument 3: path to the directory where the quantifications are stored. Right now, this is configured to be in salmon's format. In this directory, the subdirectories are the same as the sample name as indicated by the first column of argument 2. Check line 32 if your directory structure does not match this.
  • Argument 4: path to the output file that will be used in the next step. See an example under the folder example_data (data_for_regression.csv)

Step 4: Run a logistic regression model

• sex_inference_model.R
• Usage:

  Rscript sex_inference_model.R {path/to/outfile/from/step/3}
  

• Currently, this script prints out where there is a discrepancies between the predicted sex and the reported sex.
• TODO: add option for outputing sex only