MITE Data Standard JSON Schema Mega-Thread

[mmzdouc]

This thread organizes all public communication with regard to the MITE Data Standard Schema.

Please add your proposals for changes of the MITE Schema to this thread. The MITE Governing Body (defined in GOVERNANCE) will meet regularly to review proposals and decide on changes based on majority vote. The decision will also be posted in this thread.

Changelog

• 02-25: Creation of thread
• 08-25: Extended for organic/inorganic co-factors (1.8.0)

[mmzdouc]

The following extension of the MITE Standard JSON Schema is proposed:

Extending the Data Standard to explicitly add one or more co-factors that are involved in the reaction.

This addition would be on a reaction-level as a list of controlled terms.

More details will be added at a later date.

[mmzdouc]

Incorporated in the schema from version 1.8.0

[mmzdouc]

I propose the following changes to the MITE Standard JSON Schema:

Add explicit non-substrates to MITE entries

Motivation

Machine leaning application on enzyme substrate or reaction predictions often lack explicit, experimentally verified negative data (i.e. no reaction on this substrate). Currently, such applications create their negative data by e.g. random sampling of the chemical space, assuming that reactions are unlikely to occur on a random reactant/substrate. This assumption may not be valid in case of very promiscuous enzymes (e.g. JamD). By allowing MITE to store a list of experimentally verified non-substrates, this would not only support machine learning applications, but also allow to store this information, which is otherwise not available in a machine-readable format.

Possible implementation

mite_schema

The non-substrates should be on an enzyme-level (similar to auxiliary_enyzmes), since it affects the whole protein, independently of the described reactions.

A possible implementation would be:

        "non-substrates": {
          "title": "Specify experimentally verified non-substrates (negative substrates).",
          "type": "array",
          "items": {
            "type": "object",
            "additionalProperties": false,
            "properties": {
              "nonsubstrate": {
                "title": "A non-substrate in SMILES format.",
                "type": "string"
              },
              "evidence": {
                "type": "object",
                "additionalProperties": false,
                "required": [
                  "evidenceCode",
                  "references"
                ],
                "properties": {
                  "evidenceCode": {
                    "title": "Evidence for enzymatic reaction and substrate specificity",
                    "type": "array",
                    "minItems": 1,
                    "items": {
                      "$ref": "#/$defs/evidence"
                    }
                  },
                  "references": {
                    "type": "array",
                    "items": {
                      "$ref": "#/$defs/citation"
                    },
                    "minItems": 1,
                    "uniqueItems": true
                  }
                }
              }
            }
          }
        },

In practice, this could look like:

"non-substrates": [
 {
  "nonsubstrate": "CCCCCCCCCCCCC"
  "evidence": {
    "evidenceCode": [
      "heterologous expression"
    ],
    "references": [
     "doi:10.xxxxxxx"
     ]
  }
 }
]

Additionally, we could use these "non-substrates" to validate the reaction SMARTS of the entry - they should be written in a way to not accept this nonsubstrate; although I am not 100% sure if this makes sense

Please let me know your thoughts @Adafede @daenuprobst @marnixmedema

[Adafede]

I love the idea, though my only concern is the limited number of such examples in the literature?

[mmzdouc]

@Adafede I agree that this data is even more rare than experimentally confirmed substrate info 😅! But it still exists; from the top of my head, I know a couple of papers that describe this info, e.g. this one, where they introduced point mutation in the substrate peptide and checked if it can be processed by the enzyme.

When I was at the VAAM Workshop in Berlin, I was talking to a bunch of people that actually have data like that but there is not really a place to put it. By providing a place and structure for such information, we would also incentivize the publication of such data; this also touches a bit on the nano-publication we discussed the other day @Adafede (e.g. putting the experimental evidence on Zenodo and cross-referencing it with its DOI in MITE)