Mercurial > repos > pmac > iterativepca
comparison R_functions/pca_helpers.R @ 0:64e75e21466e draft default tip
Uploaded
author | pmac |
---|---|
date | Wed, 01 Jun 2016 03:38:39 -0400 |
parents | |
children |
comparison
equal
deleted
inserted
replaced
-1:000000000000 | 0:64e75e21466e |
---|---|
1 library(flashpcaR) | |
2 | |
3 # read in data from either a numeric ped file or an rds object | |
4 # output a numeric ped file, with the rownames set to the ids of the samples | |
5 get_source_data = function(data_source, data_type) { | |
6 data_type = tolower(data_type) | |
7 if (data_type == "numeric_ped") { | |
8 # read in ped file | |
9 ped_data = read.table(data_source, sep="\t", row.names=1) | |
10 } else if (data_type == "rds") { | |
11 hapmap3_object = readRDS(data_source) | |
12 ped_data = hapmap3_object$bed | |
13 } else if (data_type == "rdata") { | |
14 hapmap3_object = load_obj(data_source) | |
15 ped_data = hapmap3_object$bed | |
16 } else { | |
17 print("Unrecognised data type, returning NULL") | |
18 ped_data = NULL | |
19 } | |
20 return(ped_data) | |
21 } | |
22 | |
23 # A function that will read in and return a single object from an RData file | |
24 # This is a workaround so the program can run without needing to know name of the object; | |
25 # however the assumption is that the RData file contains only ONE object (the one we want) | |
26 load_obj = function(filename) { | |
27 # create new environment | |
28 env = new.env() | |
29 # load the rdata file into the new environment, and get the NAME | |
30 # of the first object | |
31 object_name = load(filename, env)[1] | |
32 # return the object | |
33 return(env[[object_name]]) | |
34 } | |
35 | |
36 # remove unwanted rows or columns (samples and snps, respectively) from | |
37 # the ped data | |
38 filter_ped_data = function(ped_data, xsamples, xsnps) { | |
39 # rows to remove | |
40 rr = which(rownames(ped_data) %in% xsamples) | |
41 # remove rejected samples | |
42 if (length(rr) != 0) { | |
43 fpd1 = ped_data[-rr, , drop=FALSE] | |
44 } else { | |
45 fpd1 = ped_data | |
46 } | |
47 # remove all zero and rejected snp columns | |
48 snps = which(colnames(ped_data) %in% xsnps) | |
49 zeros = which(colSums(abs(fpd1)) == 0) | |
50 cr = union(snps, zeros) | |
51 if (length(cr) != 0) { | |
52 fpd2 = fpd1[, -cr, drop=FALSE] | |
53 } else { | |
54 fpd2 = fpd1 | |
55 } | |
56 # remove monomorphic snps | |
57 snp_sds = apply(fpd2, 2, sd) | |
58 clean_ped = fpd2[, snp_sds >= 0.01, drop=FALSE] | |
59 return(clean_ped) | |
60 } | |
61 | |
62 # Ethnicity file requirements: | |
63 # - tab delimited | |
64 # - Must have at least two columns | |
65 # - First column has sample ID's | |
66 # - Second column has ethnicities | |
67 # - First row must be a header | |
68 parse_ethnicity_file = function(eth_filename) { | |
69 if (file.exists(eth_filename) == FALSE) { | |
70 print(paste0("Warning: Ethnicity file: ", eth_filename, " not found")) | |
71 return(NULL) | |
72 } | |
73 if (file.info(eth_filename)$size == 0) { | |
74 print(paste0("Warning: Ethnicity file: '", eth_filename, "' is empty")) | |
75 return(NULL) | |
76 } | |
77 eth_data = read.table(eth_filename, header=TRUE, sep="\t") | |
78 rownames(eth_data) = eth_data[, 1] | |
79 colnames(eth_data)[1] = "IID" | |
80 colnames(eth_data)[2] = "population" | |
81 return(eth_data) | |
82 } | |
83 | |
84 # Read in a file and return the first column as a | |
85 # character vector | |
86 get_first_column = function(fname) { | |
87 rv = c() | |
88 if (file.exists(fname) == FALSE) { | |
89 print(paste0("Warning: File: '", fname, "' not found")) | |
90 return(rv) | |
91 } | |
92 if (file.info(fname)$size == 0) { | |
93 print(paste0("Warning: File: '", fname, "' is empty")) | |
94 return(rv) | |
95 } else { | |
96 rv = as.character(read.table(fname)[, 1]) | |
97 return(rv) | |
98 } | |
99 } | |
100 | |
101 # Do pca using flashpcar. Returns a 2 element list | |
102 # values - contains the loadings of the pcs | |
103 # Will be an n x m matrix, where | |
104 # - n = Number of samples | |
105 # - m = number of pcs | |
106 # ids - Character array of ids, same length as number of rows in values | |
107 do_pca = function(ped_data) { | |
108 pca_data = list() | |
109 pm = data.matrix(ped_data) | |
110 values = flashpca(pm, ndim=6)$vectors | |
111 pca_data$values = values | |
112 pca_data$ids = as.character(rownames(ped_data)) | |
113 return(pca_data) | |
114 } |