Closed andrewjmc closed 4 years ago
This is probably due to file 256 having very few clustered MS2 spectra in common with these runs. The easy solution is to require at least, say, 250 retention time pairs and not consider the alignment otherwise. I'll add this check and create a new binary.
In the long run it would, however, be better to come up with a scheme that penalizes the RMSE of alignments with very few retention time pairs.
Super, that's very helpful. It seems I am quite a destructive software tester!
Now I get a different error with this batch of files. Tested twice:
Aligned runs: 0 249: rmseComb = 4.31576 rmse1 = 2.03567 rmse2 = 3.8055
Aligned runs: 0 250: rmseComb = 4.32356 rmse1 = 2.13522 rmse2 = 3.75953
Aligned runs: 0 251: rmseComb = 1.09983 rmse1 = 0.225167 rmse2 = 1.07654
Aligned runs: 0 252: rmseComb = 1.10718 rmse1 = 0.259187 rmse2 = 1.07641
Aligned runs: 0 253: rmseComb = 0.919413 rmse1 = 0.372639 rmse2 = 0.840512
Aligned runs: 0 254: rmseComb = 3.53938 rmse1 = 1.24245 rmse2 = 3.31414
Aligned runs: 0 255: rmseComb = 2.25529 rmse1 = 1.20492 rmse2 = 1.90643
Unhandled Exception: System.AccessViolationException: Attempted to read or write protected memory. This is often an indication that other memory is corrupt.
at mainCRTStartup()
That could be related to the previous error. Please try the new binary installer: quandenser-v0-02-win64-msvc-x64.zip
Rerunning now and have passed the 0 255 alignment without error, thanks!
Pending the rerun with the new version from #12, I set up a new run with 296 samples (most were also in the previous run).
Run alignment proceeded up to 24/255 then there was an error:
Exception caught: invalid vector<T> subscript
which I presume comes from 24/256.23/256 proceeded fine, though I note that standard errors were unusually small:
The same seems to apply to some other runs involving 256 too (though not all):
I tried to tee the log to a file in PowerShell but it only seems to have got the Dinosaur log. I've copied as much of the console as I can: log2.txt
The mzML for file 256 has the third smallest feature file (10,450 kB) but with a full range of retention times. File 24 seems fine.
I'll run the new binary with the old set of files first to see if original error is bypassed. Then I'll also run it with the new set and see if anything changes.
Best wishes,
Andrew