Possible hash mismatch?

berlinguyinca / spectra-hash

The splash, this is the reference documentation

http://splash.fiehnlab.ucdavis.edu

BSD 3-Clause "New" or "Revised" License

23 stars 11 forks source link

Possible hash mismatch? #29

Closed egonw closed 7 years ago

egonw commented 7 years ago

Fwd from @johnmay on G+: https://plus.google.com/u/0/+JohnMay/posts/3kTdkBhripx?sfc=true

"Sounds neat... but just found this on HMDB on first try. Others do hash to
the same so it does work just maybe has false negatives which is undesirable
for a hashcode (FP=Okay)?

Spectrum in HMDB: http://www.hmdb.ca/spectra/c_ms/2161
Spectrum in MONA: http://mona.fiehnlab.ucdavis.edu/spectra/display/HMDB00010_2161

[...]"

It's for the same compound, looks like the same spectrum, but different splashes...

ssmehta commented 7 years ago

Thanks for passing this along, Egon. This seems to be due to the fact that HMDB retains 0 intensity fragments, whereas MoNA removes upon importing. I had forgotten we had decided to retain these fragments when computing the SPLASH.

For the sake of keeping the SPLASH consistent, we will also store zero ions in MoNA. I'll have this updated tomorrow.

ssmehta commented 7 years ago

HMDB spectra have been updated, and the SPLASHes now match between HMDB and MoNA. I've also verified that none of our other data sources include zero-intensity ions.