Align kcat assignment hierarchy for reactions with assigned EC numbers

[ZyklotronAI]

Hi GECKO developers and community,

I’ve been exploring the kcat assignment process in GECKO and wanted to share and discuss my thoughts regarding reactions that lack gene associations.

From my understanding, GECKO tries to assign kcat values to these reactions using the getStandardKcat function, which is intended to return the median kcat of the corresponding subsystem, or the global median if no sufficient subsystem-specific values are available. However, when an EC number is available for such a reaction, the fuzzyKcatMatching function is not used—even though it serves as the primary method for retrieving kcat values from BRENDA for gene-associated reactions.

To improve consistency across the model, GECKO could apply the same fuzzyKcatMatching logic for any reaction that has an EC number, regardless of gene assignment. Since fuzzyKcatMatching retrieves kcat values based on the closest related organism from BRENDA (and often not the correct organism even when a gene is assigned), it seems to me like a reasonable first step for reactions without gene assignments before falling back on median-based estimates.

The corresponding molecular weights could either be calculated as the median value from all model MW values, as is currently the case for the standard kcat values. Another idea I had was that it could be estimated from enzymes with the same function from other organisms, as molecular weight is sometimes even used for cross-species protein identification, and I therefore see a good chance that it has a higher probability for valid MW prediction than taking the median value.

I’d be interested to hear your perspective on this, and whether there are specific reasons for the current approach that I might be missing. If this has come up before, I’d be grateful if you could kindly let me know or share where I can find the discussion.

Thanks for your valuable work on GECKO!