Closed shuzhao-li closed 1 year ago
Funny no error in v1.11.4
:asari-metabolomics-1.11.4 lish$ time python3 -m asari.main process --input /Users/lish/data/VAX005/C18pos --output /Users/lish/data/VAX005/asaripos
~~~~~~~ Hello from Asari (1.11.4) ~~~~~~~~~
Working on ~~ /Users/lish/data/VAX005/C18pos ~~
~~~~~~ Got 11744 khipus, with 30262 features ~~~~~~~
Annotation of 69314 Empirical compounds was written to /Users/lish/data/VAX005/asaripos_asari_project_7793041/Feature_annotation.tsv.
Feature table (84042 x 1204) was written to /Users/lish/data/VAX005/asaripos_asari_project_7793041/export/full_Feature_table.tsv.
Filtered Feature table (16812 x 1204) was written to /Users/lish/data/VAX005/asaripos_asari_project_7793041/preferred_Feature_table.tsv.
Unique compound table (13417 x 1204) was written to /Users/lish/data/VAX005/asaripos_asari_project_7793041/export/unique_compound__Feature_table.tsv.
Removing temporary pickle files...
real 53m53.150s
user 114m57.110s
sys 2m37.132s
Okay, I think there are at least two issues here.
First, I made the assumption that the default value of iter was 1 for lowess. Upon reviewing the source code for statsmodels.nonparametric.smoothers_lowess, that value should be 3. I will fix that now in default parameters.
Second, since we are passing the fraction parameter to lowess, the subset of peaks selected for a regression may not be the same between runs. So the error could be intermittent too.
Since reverting the code fixed the issue and the default argument has been updated and this issue has not recurred I'm going to close the issue.