search

DivyaratanPopli / Kinship_Inference

This is a tool to estimate pairwise relatedness from ancient DNA, taking in account contamination, ROH, ascertainment bias.
GNU General Public License v3.0
7 stars 2 forks source link

Warning running KINgaroo #21

Closed dorozco1504 closed 3 months ago

dorozco1504 commented 3 months ago

Hi, I'm having this warning at (what I think it is) the end of the KINgaroo step. It looks like this:

Creating input files from chromosome 20...
Creating input files from chromosome 21...
Creating input files from chromosome 22...
Finished merging all chromosomes...
Estimating p_0...
Inferring positions of long ROH in E13_rnchr
Inferring positions of long ROH in E18_rnchr
Inferring positions of long ROH in E19_rnchr
Inferring positions of long ROH in E3_rnchr
Inferring positions of long ROH in E4_rnchr
Inferring positions of long ROH in E5_rnchr
Inferring positions of long ROH in E7_rnchr
Inferring positions of long ROH in E8_rnchr
/data/users/dorozco/.conda/envs/kin-3.1.3/lib/python3.8/site-packages/KINgaroo/KINgaroo_scripts/hbd_hmm_functions.py:355: RuntimeWarning: invalid value encountered in long_scalars
  goodprop=np.sum((harr>=0.7) & (harr<1.1))/np.sum(harr<=1.1)
/data/users/dorozco/.conda/envs/kin-3.1.3/lib/python3.8/site-packages/KINgaroo/KINgaroo_scripts/hbd_hmm_functions.py:355: RuntimeWarning: invalid value encountered in long_scalars
  goodprop=np.sum((harr>=0.7) & (harr<1.1))/np.sum(harr<=1.1)

Do you have any idea of what's going on? Should I trust the output? I ended up with all this files:

goodpairs.csv   identical_overlap.csv    interval.txt
bedfiles              gw_sites.bed    input_diffs_hmm.csv      overlap.csv
chrm_list.csv         hapProbs        input_hbd_hmm_diffs.csv  screenlog.0
filtered_windows.txt  hbd_results     input_hbd_hmm_total.csv  splitbams
goodlibraries.csv     hmm_parameters  input_total_hmm.csv      target_samples.txt

Thanks!

DivyaratanPopli commented 3 months ago

Hi, it looks like some of the libraries have very low-coverage data, so the script can not find any genimoc windows with enough data to calculate ROH probability. You can trust the output though, by default such low-coverage libraries are assigned 0 probability of ROH, and the relatedness results with KIN will just give you low-confidence results (log likelihood<1) for these library pairs.