YFitter Files

############################################################################### 
##                                                                        
## Yfitter - a program for assigning haplogroups using maximum likelihood 
##
##

### Overview ###

Yfitter is a program for assigning Y chromosome haplogroups to individuals sequenced at low coverage. It uses a dynamic programming approach to find the maximum likelihood haplogroup given the reads observed, and is designed to be used in a samtools/bcftools pipeline. Yfitter also supports haplogrouping using chip genotype data.

This program is written by Luke Jostins (lj4@sanger.ac.uk), to whom bugs should be reported. It is destributed under the GPLv3 lincense (see LICENSE file for details). 


### Installation ###

You can compile the Yfitter program using GCC (or your C++ compiler of choice):

> g++ -o Yfitter Yfitter.cpp


### Usage instructions ###

Usage:

  Yfitter [-m -s] [-q Q] tree.xml likelihoods.qcall

The likelihoods should be in QCall format (such as produced by bcftools). Output consists of the major haplogroup followed by the highest resolution haplogroup.

Use "-m" if there is more than one sample in the file; output is one line per sample, with the first field containing the sample ID, followed by the major haplogroup, the maximum likelihood haplogroup (best guess) and confidence haplogroup (conservative guess). 

The "-q" argument sets the difference in log likelihood that defines the confidence haplogroup (make this larger if you want a more conservative confidence haplogroup assignment). By default, this is 8.685, corresponding to a deltaAIC of 4.

Use -s to print negative log likelihoods for all major haplogroups. This is useful for QC - if no haplogroup has significantly lower score than others, this suggests that haplogrouping has not been very successful.


Usage with Samtools/BCFTools:

You can generate haplogroup information for a number of BAM files using samtools to calculate likelihoods, bcftools to generate a QCall file, and Yfitter to assign haplogroups to each sample.

> samtools view -uf reference.fa sam1.bam sam2.bam [...] > allYs.bcf
> bcftools view ./karafet_sites_b36.pos allYs.bcf > infosites.qcall
> Yfitter -m karafet_tree_b36.xml infosites.qcall


Usage with Plink:

You can also use plink files. The Yfitter package includes a python script, tped2qcall.py, that converts plink .tped/.tfam files into qcall files ready for haplogrouping. For instance, if you have a .bed/.bim/.fam file set, you can extract the Y chromosome as a tped, convert this to qcall format, and use this for haplogrouping, like so:

> plink --bfile infile --chr Y --recode --transpose --out outfile
> python tped2qcall.py outfile > outfile.qcall
> Yfitter -m -q 1 karafet_tree_b36.xml outfile.qcall

Note that the tped2qcall.py script sets the negative log likelihoods to 0 for the true genotype, and 1 to for incorrect genotypes, so the -q argument sets the confidence haplogroup as the common ancester of all haplogroup with 1 more mismatchs than the best guess haplogroup (i.e. allowing for 1 genotyping error).


### The tree file ###

Yfitter requires a haplogroup tree with mutations on it to assign haplogroups. These haplogroup trees are specified in phyloXML format, with mutations on the branches specificed with the "propery" tag like this:

 <property datatype="xsd:string" ref="point_mutation:MUTNAME" applies_to="parent_branch">Y:POS,A,B</property>

Where MUTNAME is the mutation ID, POS is the chromosome position on the Y chromosome, A is the ancestrual allele and B is the derived allele.

The karafet_tree_b36.xml (and *_b37.xml) tree contains 439 mutations from Karafet et al (2008), mapped to build 37 and build 37 respectively. These consist of all non-G/C or A/T SNPs that were mapped uniquly. A few mutations have been removed since they have sinced been mapped by dbSNP to autosomes.