search

WGLab / LIQA

Long-read Isoform Quantification and Analysis
Other
39 stars 13 forks source link

LIQA quantify crashes with newest version of lifelines and -f_weight 0 #23

Closed jpfry327 closed 6 months ago

jpfry327 commented 2 years ago

If I run liqa -task quantify -refgene test.refgene -bam test.bam -out tmp.out -max_distance 10 -f_weight 0 given the files here: liqa_test.zip, I get the error:

Traceback (most recent call last):
  File "/panfs/home/yang4414/fryxx094/data/long_reads/encode/.snakemake/conda/260abc8a7999313925724c944b07e7e7/lib/python3.9/site-packag
    raise ValueError("Empty array/Series passed in.")
ValueError: Empty array/Series passed in.

During handling of the above exception, another exception occurred:

Traceback (most recent call last):
  File "/panfs/home/yang4414/fryxx094/data/long_reads/encode/.snakemake/conda/260abc8a7999313925724c944b07e7e7/lib/python3.9/site-packag
    kmf.fit(Tforkmf, event_observed=Eforkmf)
  File "/panfs/home/yang4414/fryxx094/data/long_reads/encode/.snakemake/conda/260abc8a7999313925724c944b07e7e7/lib/python3.9/site-packag
    return function(model, *args, **kwargs)
  File "/panfs/home/yang4414/fryxx094/data/long_reads/encode/.snakemake/conda/260abc8a7999313925724c944b07e7e7/lib/python3.9/site-packag
    return self._fit(durations, event_observed, timeline, entry, label, alpha, ci_labels, weights)
  File "/panfs/home/yang4414/fryxx094/data/long_reads/encode/.snakemake/conda/260abc8a7999313925724c944b07e7e7/lib/python3.9/site-packag
    (self.durations, self.event_observed, self.timeline, self.entry, self.event_table, self.weights) = _preprocess_inputs(
  File "/panfs/home/yang4414/fryxx094/data/long_reads/encode/.snakemake/conda/260abc8a7999313925724c944b07e7e7/lib/python3.9/site-packag
    durations = np.asarray(pass_for_numeric_dtypes_or_raise_array(durations)).reshape((n,))
  File "/panfs/home/yang4414/fryxx094/data/long_reads/encode/.snakemake/conda/260abc8a7999313925724c944b07e7e7/lib/python3.9/site-packag
    raise ValueError("Values must be numeric: no strings, datetimes, objects, etc.")
ValueError: Values must be numeric: no strings, datetimes, objects, etc.

I only get this error after I updated the lifelines package to version v0.27.1. I get the expected output using v0.26.4 with no error.

I'm using LIQA on very sparse data (sometimes 1 or 2 reads), which causes the modeling to break in some circumstances. I think that's what's happening here. In earlier versions of lifelines I think they passed on this error, but in the new version they may have explicitly raised the error, which we are seeing now. That's at least my guess.

agouru55 commented 6 months ago

I've run the same commands on the test files you have attached, and you are likely correct that with the updated version of lifelines, this error is passed. I would keep your environment with this version of lifelines (0.26.4) to ensure LIQA runs optimally.