KeyError: "Can't open attribute

[690645258]

Hi,

I just tried demultiplexing a run with the following command:

poreplex -i fast5_pass -o demultiplexed --trim-adapter --barcoding --fast5

Which failed due to an unknown error according to the output:

`==== Result Summary ====
BC1 BC2 BC3 BC4 undetermined
Processing failed

• File could not be opened due to unknown error 8554 30492 819 33735 22579
• Signal is too short 0 0 0 0 318
• Signal scaling QC failed 0 0 0 0 3`

Looking at the log files, the following error is coming up for each read:

2020-02-27 18:20:43,004 [signal_analyzer.py:114] (FAL30205_pass_2b4b39c4_108.fast5#0076af39-aa83-4ff6-8a27-ef145447359b) Unhandled exception KeyError: "Can't open attribute (can't locate attribute: 'num_events_template')" Traceback (most recent call last): File "/Users/yan/opt/miniconda3/lib/python3.7/site-packages/poreplex/signal_analyzer.py", line 114, in process siganal.process() File "/Users/yan/opt/miniconda3/lib/python3.7/site-packages/poreplex/signal_analyzer.py", line 259, in process events = self.load_events() File "/Users/yan/opt/miniconda3/lib/python3.7/site-packages/poreplex/signal_analyzer.py", line 315, in load_events events = self.npread.load_fast5_events() File "/Users/yan/opt/miniconda3/lib/python3.7/site-packages/poreplex/signal_loader.py", line 265, in load_fast5_events bcall = self.fast5.get_basecall() File "/Users/yan/opt/miniconda3/lib/python3.7/site-packages/poreplex/fast5_file.py", line 142, in get_basecall summary['num_events'] = int(segattrs['num_events_template']) File "h5py/_objects.pyx", line 54, in h5py._objects.with_phil.wrapper File "h5py/_objects.pyx", line 55, in h5py._objects.with_phil.wrapper File "/Users/yan/opt/miniconda3/lib/python3.7/site-packages/h5py/_hl/attrs.py", line 60, in __getitem__ attr = h5a.open(self._id, self._e(name)) File "h5py/_objects.pyx", line 54, in h5py._objects.with_phil.wrapper File "h5py/_objects.pyx", line 55, in h5py._objects.with_phil.wrapper File "h5py/h5a.pyx", line 77, in h5py.h5a.open KeyError: "Can't open attribute (can't locate attribute: 'num_events_template')"

Any ideas on what went wrong or have oxford changed the fast5 file structure and broke poreplex because of it?

[690645258]

Managed a fix for myself by modifing fast5_file.py in a couple places by making sure I used the guppy output fast5, not original.

The changes to fast5_file.py were:

groupno = analyses.name.rsplit('_', 1)[1] #ensures selection of Basecall_1D_001 and Segmentation_001 for any extra analysis

summary['num_events'] = int(len(fastqenc[1])) #fixed my original error, brought up another issue where it couldnt find Events or Move, fixed with above change to groupno

[matthew-valentine]

I was facing the same error you got originally and edited the fast_file.py file as you suggested, however I am still getting the secondary Neither Events or Move table found in the basecall error. Are those definitely the only two places you modded the file (though it's been a long time I'm sure you can't remember!)?

[690645258]

Those are all the modifications I remember doing, I think I used guppy to rebasecall and the --fast5_out and ran it on those fresh fast5's. Might be worth digging into your fast5s and check if analyses.name.rsplit('_', 1)[1] is pointing to the right things.

I only ended up using poreplex for one run, and now I've swapped to https://github.com/Psy-Fer/deeplexicon which is still actively maintained. Its slower than poreplex but works reliably and at this point is what I'd recommend.

[matthew-valentine]

Thanks for getting back to me. I am currently rerunning guppy with --fast5_out, so I will give it a go on the fast5s I get from that. I'll also have a look at deeplexicon. I was put off by the fact I would have to retrain it for our barcodes, as we used the barcodes that Poreplex was built for, but hopefully it won't be too arduous!