Better handling of -999 values and ranges

matthiaskoenig / brendapy

BRENDA parser in python

GNU Lesser General Public License v3.0

19 stars 6 forks source link

Better handling of -999 values and ranges #38

Open matthiaskoenig opened 4 years ago

matthiaskoenig commented 4 years ago

Updated the issue to reflect the discussion below.

Hrovatin commented 3 years ago

Is this related to errors like:

ERROR:root:data could not be converted to float: 2-8 {NADH} ERROR:root:data could not be converted to float: -999 {more} # Sometimes also associated with metabolite name

What do -999 for rate and {more} instead of metabolite mean mean?

When will this be fixed approx?

DeepaMahm commented 3 years ago

@Hrovatin

999 means there is no value reported for that corresponding entry (metabolite concentration or Km or Vmax). Hope this helps!

Hrovatin commented 3 years ago

Thank you very much. Do you know if any fixes that will solve these errors are planned?

matthiaskoenig commented 3 years ago

@Hrovatin This are issues with the Brenda entries. I.e. the entries exist in the database, but have no numerical values. You can easily filter these out or I could provide an option to filter these. I will have a look asap, but pretty busy today and tomorrow with other things.

Hrovatin commented 3 years ago

I would appreciate if you made a fix for the following: I think there are 2 cases:

-999 is as said by @DeepaMahm a filler for a missing value. So instead of an error you could output logger Info that no value was found or something similar.
a range (e.g. 2-8) is a valid value and I do not want to filter this out. In this case a tuple of 2 floats instead of single float could be returned (or some similar solution to represent the range).

matthiaskoenig commented 3 years ago

The reasons I kept the -999 entries was because of the associated references which could provide valuable information to modelers. I.e. here is a publication about the kinetics or parameters of this enzyme. The solution I prefer is to

clearly indicate these are missing values via NaN which would also make it very easy to filter these entries
the tuple idea is great
I would add an additional type attribute which would be MissingValue, Value, or Range which would make it easy to distinguish the different types programatically.

Hrovatin commented 3 years ago

That sounds great.

Hrovatin commented 2 years ago

Have these changes been implemented already?