UPDetector

Code and input parameters to classify putative cases of uniparental disomy (UPD) based on runs of homozygosity (ROH)

Contact: priyanka_nakka@brown.edu

HOW TO USE:

This program uses ROH-based statistics to classify putative cases of UPD using a logistic regression framework. It requires 4 input files:

1. One of the 5 classifier parameters files given here. Each file corresponds to model parameters for UPD classifiers trained in a different population; please choose the population that most closely matches that of your dataset: NE_classifier_parameters.txt - model parameters for northern European population cohorts

   SE_classifier_parameters.txt - model parameters for southern European population cohorts

   LAT_classifier_parameters.txt - model parameters for Latino/a population cohorts

   AA_classifier_parameters.txt - model parameters for African American population cohorts

   EA_classifier_parameters.txt - model parameters for East Asian population cohorts

2. Files of tab-delimited ROH-based metrics for each individual in your dataset, one for each chromosome (1-23). The file names should be in the format ‘prefix’ + chromosome number (ex. ./example-input.chr1). The first column should contain individual ids for every individual in your dataset followed by two columns of ROH-based metrics. The two metrics that we trained the classifier on are total Class C ROH length for the given chromosome in base pairs (second column) and the ratio of the total class C ROH length for the chromosome with the second largest total class C ROH length to the maximum ROH length over the all the chromosomes (third column). An example file is shown below.

    ids	    totclen		secondmax

    IND1	236000000	0.05

    IND2	60000		0.95

            ... 

3. Prefix for output files, one for each chromosome, that will contain probabilities of being a UPD case for every individual for every chromosome (ex. ./example-out1)

4. Output file name for final list of putative UPD cases