ADMIXTOOLS 2 can read genotype data in three formats.
PACKEDANCESTRYMAP and EIGENSTRAT are described here
In all three formats a dataset consists of three files: one file for the genotype matrix, one file for the SNP metadata, and one file for the sample metadata.
The following shows what a data set for three SNPs and five samples grouped into two populations (French.DG, Onge.DG) could look like:
PLINK .fam files, have six columns, but only the first
two (family/population ID and sample ID) are relevant in ADMIXTOOLS
2
French.DG S_French-1.DG 0 0 M -9
French.DG S_French-2.DG 0 0 F -9
French.DG B_French-3.DG 0 0 M -9
Onge.DG BR_Onge-2.DG 0 0 F -9
Onge.DG BR_Onge-1.DG 0 0 F -9PLINK .bim file:
1 rs3094315 0.02013 752566 G A
1 rs12124819 0.020242 776546 A G
1 rs28765502 0.022137 832918 T CEIGENSTRAT and PACKEDANCESTRYMAP use the same format for the metadata. The only difference is that the genotype data is stored in a text file in EIGENSTRAT, but in a binary file in PACKEDANCESTRYMAP.
The .ind file has 3 out of the 6 columns in the
.fam file, in different order:
S_French-1.DG M French.DG
S_French-2.DG F French.DG
B_French-3.DG M French.DG
BR_Onge-2.DG F Onge.DG
BR_Onge-1.DG F Onge.DG.snp files are the same as .bim files,
except that the first two columns are swapped:
rs3094315 1 0.02013 752566 G A
rs12124819 1 0.020242 776546 A G
rs28765502 1 0.022137 832918 T CEIGENSTRAT .geno files look like this:
29101
10211
22901One SNP per row, one sample per column, and missing data is denoted
by 9.
While there are other R packages for reading and writing binary PLINK files (plink2R, genio), no such packages exist to date for PACKEDANCESTRYMAP format files, so some of the functions shown here may be useful outside of the ADMIXTOOLS framework.
Note that the functions below impose the restriction than individual IDs may not be duplicated across populations or families.
geno = read_plink("plink/prefix")
geno = read_eigenstrat("eigenstrat/prefix")
geno = read_packedancestrymap("packedancestrymap/prefix")
# read SNPs 1000 to 2000
geno = read_packedancestrymap("packedancestrymap/prefix", first = 1000, last = 2000)
# read only ind1 and ind2
geno = read_packedancestrymap("packedancestrymap/prefix", inds = c('ind1', 'ind2'))Populations can be extracted from PLINK files (and written to another set of PLINK files) like this:
extract_samples("input/prefix", "output/prefix", pops = c("pop1", "pop2"))extract_samples can also be used to extract samples,
similar to plink --keep.
ADMIXTOOLS 2 can’t merge data sets. Data in PLINK
format can be merged using plink --bmerge.
afs = plink_to_afs("plink/prefix")
afs = eigenstrat_to_afs("eigenstrat/prefix")
afs = packedancestrymap_to_afs("packedancestrymap/prefix")This function converts genotype files from PACKEDANCESTRYMAP
or EIGENSTRAT to PLINK format. If inds or
pops is specified, only those samples will be
extracted.
ADMIXTOOLS 2 currently doesn’t provide the option to write binary PACKEDANCESTRYMAP files, so the conversion only goes one way.
eigenstrat_to_plink("eigenstrat_input_prefix", "plink_output_prefix")
packedancestrymap_to_plink("packedancestrymap_input_prefix", "plink_output_prefix")There are several ways in which admixture graphs can be represented. Some of them are useful because they can be stored as human readable text files, and others are useful because they make it easier for R to process graphs. ADMIXTOOLS 2 has several function for importing, exporting and converting between formats.
Before fitting an admixture graph, the graph only has nodes and edges. After fitting a graph, each edge has a weight associated with it, and maybe some other information. Here I mostly show examples of graphs that are not fitted yet.
The simplest way to represent a graph in a text file is by writing each edge in the graph in a separate line:
R O
R n1
n1 n2
n1 n3
n2 A
n2 admix
n3 B
n3 n4
n4 C
n4 admix
admix DWe can save this in a text file graph.txt and read it
into R like this:
graph_el = read_table2('~/Downloads/graph.tsv', col_names = F)
plotly_graph(graph_el)qpgraph() uses this format to represent fitted graphs.
In addition to the first two columns (from and
to), there will be additional columns to indicate the type
of edge, its estimated weight, and possibly more information.
## # A tibble: 6 × 6
## from to type weight low high
## <chr> <chr> <chr> <dbl> <dbl> <dbl>
## 1 R Chimp.REF edge 0.0724 NA NA
## 2 R N4N edge 0.0724 NA NA
## 3 N0N Altai_Neanderthal.DG edge 0.00321 NA NA
## 4 N0N N3N0 edge 0.0111 NA NA
## 5 N1N N0N edge 0.0841 NA NA
## 6 N1N N3N1 edge 0.0362 NA NAMany ADMIXTOOLS 2 functions use the igraph
package, which has its own graph format. The most important
ADMIXTOOLS 2 functions work with graphs in either the edge list
or the igraph format, but some only work with graphs in the
igraph format. It’s easy to convert between the two
formats:
class(graph_el)## [1] "tbl_df" "tbl" "data.frame"
graph = edges_to_igraph(graph_el)
class(graph)## [1] "igraph"
graph_el2 = as.data.frame(igraph::as_edgelist(graph))
class(graph_el2)## [1] "data.frame"The text representation of the graph above in the original ADMIXTOOLS software would look like this:
vertex R
vertex O
vertex n1
vertex n2
vertex n3
vertex A
vertex admix
vertex B
vertex n4
vertex C
vertex D
label O O
label A A
label B B
label C C
label D D
edge R_O R O
edge R_n1 R n1
edge n1_n2 n1 n2
edge n1_n3 n1 n3
edge n2_A n2 A
edge n3_B n3 B
edge n3_n4 n3 n4
edge n4_C n4 C
edge admix_D admix D
admix admix n2 n4If this is stored in a file called graph.txt, it can be
imported into R like this:
graph_el = parse_qpgraph_graphfile('graph.txt')