trvinh/admix

an admixture analysis tool for Python that supports raw data from 23andme, AncestryDNA, etc.

Admix

Admix is a simple tool to calculate ancestry composition (admixture proportions) from SNP raw data provided by various DNA testing vendors (such as 23andme and AncestryDNA).

• Installation
• Usage
• Output Example
• Raw Data Format
• Models
• Implementation

Installation

Install from Github

You can use pip to install Admix directly from this Github repository:

pip install git+https://github.com/stevenliuyi/admix

Install from PyPI

You can also install Admix from PyPI:

pip install admix

Note that due to the size limit, the package on PyPI only contains five models (K7b, K12b, globe13, world9 and E11). If you want all models, you could download them or just install Admix from this repository as shown above.

Usage

Suppose that you've already had your 23andme raw data downloaded and placed in the current directory with the name my_raw_data.txt. Then you can perform admixture calculation by specifying the calculation model (K7b in this example):

admix -f my_raw_data.txt -v 23andme -m K7b

You can also set multiple models for calculation:

admix -f my_raw_data.txt -v 23andme -m K7b K12b

If no models are set, the program will apply all the available models:

admix -f my_raw_data.txt -v 23andme

You can choose the raw data format by changing the -v or --vendor parameter. The values supported are listed here.

You may also set the -o or --output parameter to write the ancestry composition results into a file:

admix -f my_raw_data.txt -v 23andme -o result.txt

If you don't have your raw data yet, you can also test the program by using a demo 23andme data file provided by the program:

admix -m world9

Chinese users may turn on the -z flag so the population would be displayed in Chinese:

admix -z -m E11

Besides, you may use --sort flag to sort the proportions and --ignore-zeros flag to display non-zero proportions only.

For more help information, you could use:

admix -h

Output Example

• English

Command: admix -m K12b

Output:

Gedrosia: 0.06%
Siberian: 3.71%
Northwest African: 0.00%
Southeast Asian: 33.43%
Atlantic Med: 0.07%
North European: 0.00%
South Asian: 0.00%
East African: 0.00%
Southwest Asian: 0.01%
East Asian: 62.72%
Caucasus: 0.00%
Sub Saharan: 0.00%

• Chinese

Command: admix -m K12b -z

Output:

格德罗西亚: 0.06%
西伯利亚: 3.71%
西北非: 0.00%
东南亚: 33.43%
大西洋地中海: 0.07%
北欧: 0.00%
南亚: 0.00%
东非: 0.00%
西南亚: 0.01%
东亚: 62.72%
高加索: 0.00%
撒哈拉以南非洲: 0.00%

Raw Data Format

Admix supports raw data formats from the following DNA testing vendors with -v or --vendor parameter:

parameter value	vendor
23andme	23andme
ancestry	AncestryDNA
ftdna	FamilyTreeDNA Family Finder
ftdna2	FamilyTreeDNA Family Finder (new format)
wegene	WeGene
myheritage	MyHeritageDNA

Models

Admix supports many publicly available admixture models. All the calculator files are properties of their authors, and are not covered by the license of this program. Links are provided which contain more information for each model.

model value	model name	source
K7b	Dodecad K7b	Link
K12b	Dodecad K12b	Link
globe13	Dodecad globe13	Link
goble10	Dodecad globe10	Link
world9	Dodecad world9	Link
Eurasia7	Dodecad Eurasia7	Link
Africa9	Dodecad Africa9	Link
weac2	Dodecad weac (West Eurasian cline) 2	Link
E11	E11	Link
K36	Eurogenes K36	Link
EUtest13	Eurogenes EUtest K13	Link
Jtest14	Eurogenes Jtest K14	Link
HarappaWorld	HarappaWorld	Link
TurkicK11	Turkic K11	Link
KurdishK10	Kurdish K10	Link
AncientNearEast13	Ancient Near East K13	Link
K7AMI	Eurogenes K7 AMI	Link
K8AMI	Eurogenes K8 AMI	Link
MDLPK27	MDLP K27	Link
puntDNAL	puntDNAL K12 Ancient World	Link
K47	LM Genetics K47	Link
K7M1	Tolan K7M1	Link
K13M2	Tolan K13M2	Link
K14M1	Tolan K14M1	Link
K18M4	Tolan K18M4	Link
K25R1	Tolan K25R1	Link
MichalK25	Michal World K25	Link

Implementation

Maximum likelihood estimation (MLE) algorithm is applied for ancestry composition calculation, and the implementation is fairly straightforward.

Let F_nk be the minor allele frequency of SNP marker n for population k, l^minor_n and l^major_n be the minor and major allele for marker n respectively, and G_ni be the allele at marker n of the individual we're interested in (i=1,2). Our goal is to find the admixture fraction q_k of the individual, which maximize the log likelihood function

where χ is the indicator function, J and j are the all-ones matrix/vector. Note that the Einstein summation convention is implied here. With the constraints 0 ≤ q_k ≤ 1 and Σ q_k = 1, we can obtain the admixture proportions q_k by applying optimization techniques.