HaploHMM - A Hidden Markov Model (HMM) Based Program for Haplotype Inference Using Identified Haplotypes and Haplotype Patterns

Transcription

1 HaploHMM - A Hidden Markov Model (HMM) Based Program for Haplotype Inference Using Identified Haplotypes and Haplotype Patterns Jihua Wu, Guo-Bo Chen, Degui Zhi, NianjunLiu, Kui Zhang 1. HaploHMM HaploHMM is a program to infer haplotypes from unrelated individuals using a set of identified haplotypes and haplotype patterns from previous studies. The program is based on an extended Hidden Markov Model. It has been shown that the program is effective for genotypes with a substantial amount of missing data. Please refer to Wu et al. (2014) for the detailed description of methods used in this program. 2. Files Included in the Package HaploHMM.exe: the executive file under windows operating system HaploHMM_Manual.pdf: the manual for HaploHMM HaploHMM_Geno.txt: an example file for genotype data HaploHMM_HapList.txt: an example file for the set of identified haplotypes HaploHMM_HapPattern.txt: an example file for identified haplotype patterns HaploHMM_res.txt: an example file for haplotypes of each individual HaploHMM_res.sample11: an example file for haplotypes of each individual sampled at the HMM iteration How to run HaploHMM HaploHMM.exe runs on the Command Promote window of Microsoft Windows operating system. Here we use Microsoft Windows 7 operating system as an example to show how to run HaploHMM. To run HaploHMM, you first need to open a Command Prompt window. There are two ways to do this. The first way is to click the Start button, click All Programs, click Accessories, and then click Command Prompt. The second way is to click Start button, type Command Prompt in the Search box, then double-click Command Prompt in the list of results.

2 After a Command Promote window is opened, change the folder to the folder that contains your program and data, then type the following command with the options followed by the specified value for each option: HaploHMM --genotype [genotype data file] --haplotype [initial haplotype file] --pattern [haplotype pattern file] --seed [seed] --burnin [number of iterations for burnin] --rounds [total number of iterations] --states [number of reference haplotypes] --sampleinterval [sample interval] --prefix [prefix of output files] The order of options can be arbitrary. However, you must specify the value for each option right after that option identifier with a space between them. Also be aware that each option is preceded by --. If you do not specify a value for an option, the program will use its default value. The brief description and the default value for each option are listed in Table 1. Using the data files provided in the package, we can infer their haplotypes based on the following command: Example 1: Command line example. HaploHMM --genotype HaploHMM_Geno.txt --haplotype HaploHMM_HapList.txt -- pattern HaploIHMM-HapPattern.txt --seed burnin 10 --rounds states errorrate prefix HaploHMM_res Table 1. The description and the default value for each option used in HaploHMM. Option Description Default value --genotype [genotype data file]: A character string for the file name of No default. genotype data. This is required by HaploHMM. --haplotype --pattern --seed --burnin --rounds [initial haplotype file]: A character string for the file name of a set of identified haplotypes. [haplotype pattern file]: A character string for the file name of the sets of identified haplotype patterns. [seed]: A positive integer for the starting seed of random number generator. [number of iterations for burnin]: A non-negative integer for the number of iterations before taking samples in the HMM sampling. [total number of iterations]: A non-negative integer for the total number of iterations in the HMM sampling No default. No default

3 --states --sampleinterval --errorrate --prefix [number of reference haplotypes]: A non-negative integer for the total number of reference haplotypes used in the HMM sampling. The value of 0 means the use of all inferred haplotypes from individuals provided from [genotype data file] and all haplotypes provided from [initial haplotype file]. [sample interval]: A no-native integer for the intermediate results. A value of 0 means no intermediate results. A value of 1 means the haplotype from each HMM iteration after the burin will be output to files. [error rate]: A number in [0, 1) to represent the error rate in the HMM. [prefix of output files]: A character string for the prefix of output file names res_hmm 4. How to prepare input files (options --genotype, --haplotype, --pattern ) HaploHMM can have three different types of input file specified by options --genotype, -- haplotype, and --pattern. These files are [genotype data file], [initial haplotype file], and [haplotype pattern file]. Only [genotype data file] is required by the program and the other two options can be omitted. [genotype data file] : --genotype This is the file for genotypes. Two alleles at a bi-allelic locus are detonated by 0 and 1, respectively. Missing allele is denoted as?. If your genotype is present-absent genotype as those described in our paper (Joo et al., 2007 and Wu, et al., 2014) and is denoted by one digit (0 absence, 1 presence, and? missing) at each locus, you can convert 1 to 1?, 0 to 0 0, and? to?? to comply the file format required by HaploHMM. The detailed format of genotype data file is as follows: First raw: the list of marker names in this genotype data file Second raw : the individual ID and genotype data First column: the individual ID (an arbitrary character string) Second column : Two columns for two alleles at a marker locus. The order of alleles does not matter. The columns should be separated by the space or the tab.

4 Example 2: [genotype data file] 2DL1 2DL2 2DS1 2DL3 3DS1 E ?? 0 1 E002 0? E ? [initial haplotype file]: --haplotype This is the file for a set of identified haplotypes. Each haplotype is listed in one row as a combination of allele symbols of 0 and 1. The missing allele is not allowed in the file. The first column is the list of marker names and must match the names given in [genotype data file]. The detailed format of [initial haplotype file] is as follows: First raw: the marker names which match the names given in [genotype data file] Second raw : haplotype and its name First column: the haplotype name Second column : each column is for an allele at a marker locus for the haplotype. The columns should be separated by the space or the tab. Example 3: [initial haplotype file] 2DL1 2DL2 2DS1 2DL3 3DS1 A A A [haplotype pattern file]: --pattern This is the file for the sets of haplotype patterns. Each set of haplotype patterns consists of a header line to indicate the involved markers and a set of haplotype patterns allowed in the program. For example, if you have a set of haplotype patterns for markers 2DS2 and 2DL2 and have observed that they are in complete negative LD. In other words, you only have two types of haplotypes 1 0 and 0 1 allowed for 2DS2 and 2DL2 but do not have other two haplotypes ( 0 0 and 1 1 ). This set of haplotype patterns can be specified in the file as follows: 2DS2 2DL2 2

5 In the first row, the first two are the names of two markers and the last number is the number of haplotype patterns allowed in the program for these two markers. The subsequent rows specify the haplotype patterns.. The detailed format of [haplotype pattern file] is as follows: First section: the first set of haplotype patterns First row: the header line First column : the name of markers involved Last column: the number of haplotype patterns that are allowed Second row...: each haplotype pattern is specified in each row; Second section..: other sets of haplotype patterns Example 4: [haplotype pattern file] 3DL DS2 2DL DS1 3DL Example 5: [haplotype pattern file] 3DL2 2DL DS2 2DL2 2DL DS1 3DL It is worth noting that all identified haplotypes in initial haplotype file must be compatible

6 with one haplotype pattern in each set of haplotype patterns. For example, the haplotype ( ) for the markers 3DL2, 2DL1, 2DS2, 2DL2, 2DL3, 2DL4, 3DS1, 3DL1 contradicts with the haplotype pattern in example 5 because the haplotype pattern (1 0 1) for the markers 2DS2, 2DL2, and 2DL3 is not allowed. So this haplotype cannot be used together with the haplotype patterns in example 5. The program checks the contradiction between haplotypes in [initial haplotype file] and haplotype patterns in [haplotype pattern file]. The program will stop and reports an error message when a contradiction is found. 5. The Formats and Interpretations of Output Files HaploHMM generates an output file with the haplotype pair of each individual. If the value from the option --sampleinterval is a positive integer, the haplotype pair of each individual from certain HMM iterations after the burin will be output to files. The prefix of the output file names is specified by the option --prefix. All output files have the same format: each row consists of the name of each individual, the name of haplotype (either hap1 or hap2 ), and one of two haplotypes of that individual. Example 6: [output file] ind1 hap ind1 hap ind2 hap ind2 hap ind3 hap ind3 hap ind4 hap ind4 hap Options for HMM sampling HaploHMM provides set of options for the HMM sampling. You can set the starting seed of the random number generator by --seed, the number of iterations before taking samples by --burnin, the number of total iterations by --rounds, the number of reference haplotypes by --states, the error rate by --errorrate, and the intermediate results by -- sampleinterval. In general, the larger values of --rounds and --states are, the more computational time are needed and the more accurate results can be obtained. For the option --rounds, a

7 number greater than 100 is sufficient in most situations. The default value of --states is 0, which means that the program will use all inferred haplotypes from individuals provided from [genotype data file] and all haplotypes provided from [initial haplotype file] as reference haplotypes. Since the computational time is proportional to the square of the number of reference haplotypes, you may want to use a smaller number and a few hundred reference haplotypes are sufficient in most situations. 7. Remarks In this section, we highlight some possible solutions for the problems that may be encountered in running HaploHMM. HaploHMM can only handle bi-allelic markers and the two alleles at each marker locus should be specified as 1 and 0. The missing data is specified as?. You cannot use other allele symbols such as A, G, C, T or other numeric or character symbols. The first lines in three input files only list the marker names in the file and the order of these names is the order that data are entered. These first lines are not the variable headers. If you have a data file with the first line as column names (variable header), you need to modify the first line. Otherwise, the program will report an error message. Both [initial haplotype file] and [haplotype pattern file] can help to reduce the computing time. Without one of these input files, the program will run longer in the presence of a large of portion of missing data. You can monitor the computing progress through the screen output. To reduce the computational time, you can use the smaller number for options --rounds and --states. Please refer to Section 6 for more details. 8. Contact information This program and related materials can be downloaded through following website: Bugs and comments should be addressed to:

8 Kui Zhang Section of Statistical Genetics Department of Biostatistics University of Alabama at Birmingham 1600 University Blvd Change Log The program is first released on June 22, References Yoo Y, Kaslow R, Tang J, Zhang K Haplotype inference for present-absent genotype data for clustered genes using identified haplotypes and haplotype patterns. Bioinformatics, 23: Jihua Wu, Guo-Bo Chen, Degui Zhi, Nianjun Liu, Kui Zhang A hidden Markov model for haplotype inference for present-absent data of clustered genes using identified haplotypes and haplotype patterns. Frontiers in Statistical Genetics. Under Review.