main

Main Directory

• Each folder in this directory was meant to be for running a separate reaction (e.g. "glucose" is for glucose degradation).
• main.py contains methods and settings that will be used by reactions running in all subdirectories. It also contains some utility methods useful for seeing what's going on and debugging.
• mod_to_neo4j_exporter.py contains method(s) to export networks from the MØD pipeline that we later loaded into a Neo4j MØDule for running graph queries (searching for autocatalytic cycles, etc.).

Custom dump file formats

Species by generation:

It was important to keep track of which species was created in which generation. .txt files with names such as reaction_name_output.txt contain exactly this. For instance, here's an excerpt from glucose_degradation_output.txt

G1	C(C(C1C(C(C(O)O1)O)O)O)O
G1	C(C1C(C(C(C(O)O1)O)O)O)O
G1	C1C(C(C(C(C(O)O1)O)O)O)O
G2	C1(C(C(C(C1CO)O)O)O)=O
G2	C1(C(C(C(C(C1)O)O)O)O)=O
G2	C(C1C(C(C(C1O)O)O)O)=O

Here, G1 means the corresponding SMILES (unique for each species, canonicalized by MØD) first appeared in generation 1. G0 would include just the initial reactants.

Neo4j Imports:

The neo4j exporter creates two kinds of files, one which shows relations between species (reactions) and others just show which species exist. The format of the rels_i.txt (where i represents the generation) files for reactions is

REACTION_ID MOLECULE_SMILES CONSUMPTION/CREATION    RULE_NAME

A description of the format is as follows:

A given set of products connected to a set of edges could be formed not by just one reaction rule but multiple rules (which involve the same reactants/products). To account for that, we added a subscript to the reaction IDs where each subscript value corresponds to a different rule. Say, a given reaction edge has 'n' rules associated with it. We create separate entries for the reaction with subscripts running from 0 to n-1

Here, REACTION_ID looks something like 16_0, where 16 is the id of the hyperedge that connects the molecules (in MØD) and the subscript 0 means it's the first rule associated with that hyperedge. It was important to make a distinction because particular set of edges could associated with amidine hydrolysis as well as some sort of tautomerisation reaction, they could each be assigned a separate subscript index for the same reaction id. MOLECULE_SMILES is simply the SMILES string associated with that molecule. CONSUMPTION/CREATION is -1 if the molecule is a reactant and is 1 if it's a product. RULE_NAME is simply the name of the reaction rule which produced it.

An example is

6_0	C#N	-1	HCN Addition to Nitriles
6_0	C#N	-1	HCN Addition to Nitriles
6_0	C(C=N)#N	1	HCN Addition to Nitriles

which represents C#N + C#N -> C(C=N)#N happening.

Note: As of now, the contents of rels_n.txt are cumulative in the sense that a file, say rels_3.txt would contain the content of rels_2.txt and rels_1.txt as well. So rels_3.txt includes reactions of G1 + G2 + G3 combined. Therefore, care should be taken when trying to obtain just the reactions of G3.

Precaution: every subsequent run, these .txt dump files will not be deleted and rewritten on their own, instead additional lines will be appended at their end. So one needs to delete old files manually.

Tautomer Cleanup

We experimented with different techniques (using RDKit/MolVS as well as Ambit-tautomer) to remove redundant tautomers but none of them were fully satisfactory. Some of the (redundant) code which we no longer employ in our workflow is still here (e.g. clean_tautomers.py, ambit_tautomer.py). Ambit is a Java library (dependent on CDK) which, to our workflow, could be pipelined using jpype. See Using_Ambit.md for notes that I wrote in mid-2020 if they seem useful to you.

TODO:

• see if there's anything else useful to explain

Future Code Improvements

• The way I wrote the "cheap hax" to toggle the Cannizarro 2 rule by setting with_formaldehyde=True or =False was terrible. This boolean was defined initially in ../rules/cannizarro2.py, to decide whether to use glucose as a "fixed aldehyde" or to use formaldehyde instead. Things can easily go wrong because of how this boolean has been accessed/mutated in other scripts like main.py or the individual reaction scripts like glucose_degradation.py.
• The way these scripts have been run up to now is by performing mod -f filename.py --- that still provides most of Python functionality but I have noticed conflicts with MOD's way of dealing with package directories and Python's native way. In newer versions, importing * from libpymod directly could reduce that.
  • A more proper way of dealing with packages would then be possible.
• A lot of what I put in main.py (back in 2020) are 'utility' methods. These could go in their own separate file.
• Several lengthy methods could be modularised better for the sake of readability. Also, I learnt much cleaner ways to do certain tasks later on that I haven't updated everywhere. (These don't affect the function of the code itself, but make it more readable.)