ASW inference - Three pulses model#

This example implements inference for the ASW population under a three pulses model of admixture, using the tracts package. Inference is performed using autosomal and X chromosome data, allowing for the specification of sex-biased admixture.

To implement this example, we use the following driver file:

samples:
  directory: ./TrioPhased/
  individual_names: [
    "NA19625","NA19700","NA19701","NA19703","NA19704","NA19707","NA19711","NA19712","NA19713","NA19818","NA19819",
    "NA19834","NA19835","NA19900","NA19901","NA19904","NA19908","NA19909","NA19913","NA19914","NA19916","NA19917",
    "NA19920","NA19921","NA19922","NA19923","NA19982","NA19984","NA20126","NA20127","NA20274","NA20276","NA20278",
    "NA20281","NA20282","NA20287","NA20289","NA20291","NA20294","NA20296","NA20298","NA20299","NA20314","NA20317",
    "NA20318","NA20320","NA20321","NA20332","NA20334","NA20339","NA20340","NA20342","NA20346","NA20348","NA20351",
    "NA20355","NA20356","NA20357","NA20359","NA20362","NA20412"]
  male_names : [
    "NA19700","NA19703","NA19711","NA19818","NA19834","NA19900","NA19904","NA19908","NA19916","NA19920",
    "NA19922","NA19982","NA19984","NA20126","NA20278","NA20281","NA20291","NA20298","NA20318","NA20340",
    "NA20342","NA20346","NA20348","NA20351","NA20356","NA20362"] #see Readme_dataprocessing.md for how this was generated
  filename_format: "{name}_{label}_final.bed"
  labels: [A, B] #If this field is omitted, 'A' and 'B' will be used by default
  chromosomes: 1-22 #The chromosomes to use for analysis. Can be specified as a list or a range
  allosomes: [X]

output_filename_format: "ASW_test_output_{label}"
log_filename: 'ASW_three_pulses.log'
output_directory: ./output_three_pulses/
verbose_log: 1
verbose_screen: 30
log_scale : True

start_params:
  t1: 10
  REUR: 0.8
  RAFR: 0.9
  REUR2: 0.2
  t2: 5
  t3: 3
  REUR_sex_bias: 0.1
  REUR2_sex_bias: 0.1
  RAFR_sex_bias: 0.1
repetitions: 3
seed: 100
maximum_iterations: 1000
unknown_labels_for_smoothing: ["UNK", "centromere","miscall"] # segments with these labels will be smoother over, that is, will be filled with neighbouring ancestries up to their midpoints.
exclude_tracts_below_cm: 2
npts : 50
#fix_parameters_from_ancestry_proportions: ['REUR2', 'RAFR', 'REUR2_sex_bias', 'RAFR_sex_bias']
output_directory: ./output_three_pulses/
ad_model_autosomes: M
ad_model_allosomes: DC

Complete results from this analysis are saved in the output directory specified in the driver file. Below, we display the optimal parameters estimated from this analysis, as well as the plots illustrating the inferred tract length distributions, compared to the observed histograms, for every source population and chromosome type (autosomes and X chromosome).

Optimal parameters#

Estimated optimal parameters#

parameter

value

REUR

0.7729121375993303

REUR_sex_bias

0.00025865307547912053

t1

8.634748344515474

RAFR

0.8841928552747739

RAFR_sex_bias

-0.004675231087385101

t2

7.226408784836712

REUR2

0.14493237980731358

REUR2_sex_bias

2.944503850343949e-05

t3

5.666278423687627

Tract length histograms#

Autosomal admixture#

African ancestry tract histogram

X chromosome admixture in females#

European ancestry tract histogram

X chromosome admixture in males#

Native American ancestry tract histogram
------------------------------------------------------------------------------------------------

Running tracts 2.0 with driver file: ASW_three_pulses.yaml

Reading data, demographic model and driver specifications...

------------------------------------------------------------------------------------------------

excluding_tracts_below set to 2.0 cM.
Ancestries: ['EUR', 'NAT', 'AFR']
Data autosome proportions: [0.19578862 0.03825495 0.76595643]
Data allosome proportions: [0.16839124 0.03818939 0.79341937]
Model parameters : ['REUR', 'REUR_sex_bias', 't1', 'RAFR', 'RAFR_sex_bias', 't2', 'REUR2', 'REUR2_sex_bias', 't3']

Multiple starting parameters were generated. These will be converted to optimizer units and used for multiple optimization runs.

Run | Starting parameters
---------------------------------------------------
  1 | [0.8, 0.1, 9.867, 0.9, 0.1, 6.789, 0.2, 0.1, 3.064]
  2 | [0.8, 0.1, 9.854, 0.9, 0.1, 6.8, 0.2, 0.1, 2.506]
  3 | [0.8, 0.1, 11.99, 0.9, 0.1, 5.966, 0.2, 0.1, 3.074]
---------------------------------------------------

Optimization run #1

-----------------------------------------------------------------------------------------
Admixture is modelled with the M model for autosomes and with the DC model for allosomes.
Optimization is performed in two steps.
Step 1 : Optimizing autosomal likelihood over parameters ['REUR', 't1', 'RAFR', 't2', 'REUR2', 't3'].
Iter.    Log-likelihood  Model parameters        Transmission
-----------------------------------------------------------------------------------------
30      , -1215.63    , array([ 0.797778   ,  0          ,  9.96533    ,  0.891678   ,  0          ,  6.94546    ,  0.198058   ,  0          ,  3.68714    ]), Autosomes
60      , -889.862    , array([ 0.796526   ,  0          ,  9.58264    ,  0.885576   ,  0          ,  7.3138     ,  0.190549   ,  0          ,  4.30489    ]), Autosomes
90      , -763.181    , array([ 0.786673   ,  0          ,  9.03934    ,  0.880502   ,  0          ,  7.57498    ,  0.166511   ,  0          ,  4.80616    ]), Autosomes
120     , -727.807    , array([ 0.786846   ,  0          ,  8.85361    ,  0.878754   ,  0          ,  7.64575    ,  0.15868    ,  0          ,  4.89178    ]), Autosomes
150     , -714.745    , array([ 0.785267   ,  0          ,  8.84535    ,  0.878385   ,  0          ,  7.67761    ,  0.154349   ,  0          ,  4.94699    ]), Autosomes
180     , -703.543    , array([ 0.783706   ,  0          ,  8.8365     ,  0.877875   ,  0          ,  7.68805    ,  0.150951   ,  0          ,  4.99474    ]), Autosomes
Step 1 completed.
----------------------------------------------------------------------------------------------------------------------------
Step 2 : Optimizing autosomal + allosomal likelihood over parameters : ['REUR_sex_bias', 'RAFR_sex_bias', 'REUR2_sex_bias'].
Non-sex-bias parameters fixed at values from previous optimization step.
Iter.    Log-likelihood  Model parameters        Transmission
----------------------------------------------------------------------------------------------------------------------------
210     , -216.981    , array([ 0.783637   , -0.00231255 ,  8.837      ,  0.87784    , -0.0096321  ,  7.68785    ,  0.150934   , -0.00136707 ,  4.99366    ]), Female allosomes
210     , -96.5433    , array([ 0.783637   , -0.00231255 ,  8.837      ,  0.87784    , -0.0096321  ,  7.68785    ,  0.150934   , -0.00136707 ,  4.99366    ]), Male allosomes
210     , -703.203    , array([ 0.783637   , -0.00231255 ,  8.837      ,  0.87784    , -0.0096321  ,  7.68785    ,  0.150934   , -0.00136707 ,  4.99366    ]), Autosomes
Step 2 completed.
----------------------------------------------------------------------------------------------------------------------------

Optimization run #2

-----------------------------------------------------------------------------------------
Admixture is modelled with the M model for autosomes and with the DC model for allosomes.
Optimization is performed in two steps.
Step 1 : Optimizing autosomal likelihood over parameters ['REUR', 't1', 'RAFR', 't2', 'REUR2', 't3'].
Iter.    Log-likelihood  Model parameters        Transmission
-----------------------------------------------------------------------------------------
30      , -1536.14    , array([ 0.798213   ,  0          ,  10.1199    ,  0.8867     ,  0          ,  6.93854    ,  0.197798   ,  0          ,  2.91645    ]), Autosomes
60      , -983.654    , array([ 0.797388   ,  0          ,  10.0798    ,  0.874671   ,  0          ,  7.69938    ,  0.18867    ,  0          ,  3.3935     ]), Autosomes
90      , -811.131    , array([ 0.79505    ,  0          ,  9.23515    ,  0.863564   ,  0          ,  7.95468    ,  0.166934   ,  0          ,  3.96751    ]), Autosomes
120     , -779.995    , array([ 0.79503    ,  0          ,  9.17526    ,  0.863572   ,  0          ,  8.03091    ,  0.161757   ,  0          ,  4.02618    ]), Autosomes
Step 1 completed.
----------------------------------------------------------------------------------------------------------------------------
Step 2 : Optimizing autosomal + allosomal likelihood over parameters : ['REUR_sex_bias', 'RAFR_sex_bias', 'REUR2_sex_bias'].
Non-sex-bias parameters fixed at values from previous optimization step.
Iter.    Log-likelihood  Model parameters        Transmission
----------------------------------------------------------------------------------------------------------------------------
150     , -215.644    , array([ 0.795049   , -0.00515769 ,  9.17703    ,  0.863557   , -0.031601   ,  8.03271    ,  0.16174    , -0.014531   ,  4.02589    ]), Female allosomes
150     , -97.6737    , array([ 0.795049   , -0.00515769 ,  9.17703    ,  0.863557   , -0.031601   ,  8.03271    ,  0.16174    , -0.014531   ,  4.02589    ]), Male allosomes
150     , -779.494    , array([ 0.795049   , -0.00515769 ,  9.17703    ,  0.863557   , -0.031601   ,  8.03271    ,  0.16174    , -0.014531   ,  4.02589    ]), Autosomes
Step 2 completed.
----------------------------------------------------------------------------------------------------------------------------

Optimization run #3

-----------------------------------------------------------------------------------------
Admixture is modelled with the M model for autosomes and with the DC model for allosomes.
Optimization is performed in two steps.
Step 1 : Optimizing autosomal likelihood over parameters ['REUR', 't1', 'RAFR', 't2', 'REUR2', 't3'].
Iter.    Log-likelihood  Model parameters        Transmission
-----------------------------------------------------------------------------------------
30      , -1344.44    , array([ 0.797258   ,  0          ,  12.0966    ,  0.893864   ,  0          ,  6.61638    ,  0.198304   ,  0          ,  3.49862    ]), Autosomes
60      , -995.334    , array([ 0.792664   ,  0          ,  11.3806    ,  0.882393   ,  0          ,  6.81238    ,  0.188567   ,  0          ,  4.27351    ]), Autosomes
90      , -821.322    , array([ 0.782795   ,  0          ,  10.2064    ,  0.878202   ,  0          ,  6.96528    ,  0.172819   ,  0          ,  5.07204    ]), Autosomes
120     , -737.571    , array([ 0.780019   ,  0          ,  9.62516    ,  0.8825     ,  0          ,  7.1452     ,  0.158982   ,  0          ,  5.38853    ]), Autosomes
150     , -673.49     , array([ 0.774047   ,  0          ,  8.65556    ,  0.884256   ,  0          ,  7.20062    ,  0.147654   ,  0          ,  5.66046    ]), Autosomes
180     , -667.882    , array([ 0.773252   ,  0          ,  8.63872    ,  0.884272   ,  0          ,  7.23006    ,  0.145465   ,  0          ,  5.66193    ]), Autosomes
210     , -666.834    , array([ 0.772967   ,  0          ,  8.63523    ,  0.884189   ,  0          ,  7.2254     ,  0.145051   ,  0          ,  5.66488    ]), Autosomes
Step 1 completed.
----------------------------------------------------------------------------------------------------------------------------
Step 2 : Optimizing autosomal + allosomal likelihood over parameters : ['REUR_sex_bias', 'RAFR_sex_bias', 'REUR2_sex_bias'].
Non-sex-bias parameters fixed at values from previous optimization step.
Iter.    Log-likelihood  Model parameters        Transmission
----------------------------------------------------------------------------------------------------------------------------
240     , -213.38     , array([ 0.772912   ,  0.000273909,  8.63475    ,  0.884193   , -0.0038701  ,  7.22641    ,  0.144932   ,  2.96945e-05,  5.66628    ]), Female allosomes
240     , -96.1323    , array([ 0.772912   ,  0.000273909,  8.63475    ,  0.884193   , -0.0038701  ,  7.22641    ,  0.144932   ,  2.96945e-05,  5.66628    ]), Male allosomes
240     , -666.538    , array([ 0.772912   ,  0.000273909,  8.63475    ,  0.884193   , -0.0038701  ,  7.22641    ,  0.144932   ,  2.96945e-05,  5.66628    ]), Autosomes
Step 2 completed.
----------------------------------------------------------------------------------------------------------------------------

---------------------------------------------------------------------------
Results from multiple optimization runs with different starting parameters:
-------------------------------------
Run |       LogLik | Found parameters
-------------------------------------
  1 |     -1016.24 | [0.7836, -0.003939, 8.837, 0.8778, -0.03301, 7.688, 0.1509, -0.004883, 4.994]
  2 |     -1092.78 | [0.795, -0.005173, 9.177, 0.8636, -0.03263, 8.033, 0.1617, -0.01536, 4.026]
  3 |     -976.034 | [0.7729, 0.0002587, 8.635, 0.8842, -0.004675, 7.226, 0.1449, 2.945e-05, 5.666]
-------------------------------------

Final parameters and corresponding likelihood:
---------------------------------------------------------------------------------------------------------------------------------------
      LogLik |         REUR REUR_sex_bias           t1         RAFR RAFR_sex_bias           t2        REUR2 REUR2_sex_bias           t3
---------------------------------------------------------------------------------------------------------------------------------------
    -976.034 |       0.7729    0.0002587        8.635       0.8842    -0.004675        7.226       0.1449    2.945e-05        5.666
---------------------------------------------------------------------------------------------------------------------------------------
Results saved to : ./output_three_pulses/

{'destination_dir': PosixPath('/home/runner/work/tracts/tracts/docs/source/auto_examples/ASW/output_three_pulses'), 'table_file': PosixPath('/home/runner/work/tracts/tracts/docs/source/auto_examples/ASW/output_three_pulses/ASW_test_output_optimal_parameters.txt')}

import sys
from pathlib import Path

sys.path.append('.')

from tracts.driver import run_tracts

script_path = Path(sys.argv[0]).resolve()

driver_filename = "ASW_three_pulses.yaml"

run_tracts(driver_filename = driver_filename, script_dir = script_path)


# Don't run the code below: for documentation purposes only.
from tracts.doc_utils import prepare_example_outputs_for_docs
prepare_example_outputs_for_docs("output_three_pulses")

Total running time of the script: (2 minutes 43.189 seconds)

Gallery generated by Sphinx-Gallery