Raise DetectionError instead of creating wrong model

[himoto]

Hi @formersbach,

Let's create a model from a text file including unregistered rule words.

Prepare test_binding.txt including:

A is binding to B --> AB

Text-to-model conversion

from biomass import Text2Model

testmodel = Text2Model("test_binding.txt")
testmodel.convert(overwrite=True)

When running the code, we will get the following error message.

DetectionError: Unregistered words in line1: A is binding to B --> AB

Maybe: 'binds'.

This works as we expected.
This is because we already have a similar rule ('bind') in the reaction rules.
To make it correctly work, we need to run the following code before converting the text:

testmodel.register_word({"bind": ["is binding to"]})

OK, let's move on to next example. Imagine that we want a model of ubiquitination.

Prepare test_ubiquitination.txt including:

A ubiquitinates B --> uB

Since there is no rule like 'ubiquitinate', here I want biomass to raise DetectionError to let users know it.

# This is what I expect.
DetectionError: Unregistered words in line1: A ubiquitinates B --> uB

However, when you run the following code:

testmodel = Text2Model("test_ubiquitination.txt")
testmodel.convert(overwrite=True)

This does NOT raise errors but creates model.
When you look into the resulting model, you will soon find the model is wrong.
For example, in name2idx/species.py:

from dataclasses import make_dataclass
from typing import Dict, List

NAMES: List[str] = [
    "AubiquitinatesB", # <-- this is not desired :(
    "uB",
]

...

It seems that this unregistered word is recognized as state transition rule.

Do you have any ideas of how we should tackle this problem?

Ideas and PR to fix this is more than welcome!

[formersbach]

Hmm I see!
I believe this is due to line 1712 and 1713 in reaction_rules.py:

                reactants = two_species[0].replace(" ", "").split("+")
                products = two_species[1].replace(" ", "").split("+")

For
A + B --> C + D
these lines are necessary to remove the spaces and split at the +.
I think this line should not be triggered if there are spaces without a + present right?
So changing to

        if len(two_species[0].split(" ") > 1 & not "+" in two_species[0]:
                reactants = two_species[0].replace(" ", "").split("+")
        if len(two_species[1].split(" ") > 1 & not "+" in two_species[1]:
                products = two_species[1].replace(" ", "").split("+")

should fix this. We could also prevent state_transition from triggering at all if the above mentioned conditions arise.
Do you see a better way?

[himoto]

Hi @formersbach, thanks for the idea!

For
A + B --> C + D
these lines are necessary to remove the spaces and split at the +.
I think this line should not be triggered if there are spaces without a + present right?

Yes, absolutely. This is exactly what I thought to implement. If there is no "+" symbol but a space inside reactants/products, it should be wrong.
Based on your idea & comment, I have raised an issue (#231) and created a pull request (#232) that attempted to close the issue.
I would be grateful if you could review the updates. When this is merged, I will release v0.11 that includes improvements on state_transition rule.

Thank you very much.