updated bioavailability_ma_et_al

[bethanyconnolly]

Updated the templates for this task to improve grammar and that this data is only for the bioavailability via oral pathway.
Note: Oral bioavailability is not really a binary task, it should be a fraction. This paper classified oral bioavailability as “positive” if its bioavailability≥ 20%, otherwise “negative” because they needed it to be binary for their SVM model to do classification.
This means that we might need to think more about this task, some options:

• Leave it as it is right now
• Use the raw data instead of the SVM binary data: this would change it to a continuous task: http://modem.ucsd.edu/adme/databases/databases_bioavailability.htm
• Change the templates to be a bit more vague but stick with the existing data e.g. instead of this SMILES 'is orally bioavailable', we could say this SMILES has 'high oral bioavailability', or 'is likely to show oral bioavailability'.
  Happy to discuss what we think is best :)

[kjappelbaum]

Wow! That's a detailed analysis 🕵🏽‍♀️

personally, I'd go with the raw data as we could then also do the regression tasks (as you said)

[jackapbutler]

Depending on how much work it is to duplicate the template we could create a new one for option 2) and change the current one to be more like 3) so it has numerical + "high / low" language views of the data.

[MicPie]

Yeah, I guess the idea from Jack would be the best approach to work with the data from the current PR and add the regression dataset as a new one.
In the templates we should only need to change
is {bioavailable#not &NULL}{bioavailable__names__adjective}
to
has a {bioavailable#low &high }{bioavailable__names__noun}.
I'll check that. :-)

[MicPie]

PR #388 adds the proposed changes from this PR.
Issue #389 covers the addition of the raw data in a new tabular dataset.