PredictLocalization issue

[DZQ339]

github issue.zip

Description of the issue:

We got a draft model from the ‘getKEGGModelForOrganism’ step, which contains 1841 reactions and 1906 metabolites.
The gene compartments were predicated using deeploc 2.1 and got the GSS score form deeploc file.
When we predicated the localizations using the predictLocalization (predictLocalization (KEGGHomology, GSS, 'Cytoplasm', 0.5, 15, true), the new outmodel contains more than 6735 reactions and 3680 metabolites.
It seems that the predictLocalization step added the transport reaction into all of predicated compartments.

Reproducing this issue:

INSERT CODE HERE

System information

I hereby confirm that I have:

• Checked that a similar issue does not already exist or is already discussed.

[xztJLAU]

Don't forget that eukaryotes may have genes on different chromosomes that produce the same transcript, let's regenerate the rules of the gene-protein -reactions（GPR）

% Generate reaction equations
rxnFormula = printRxnFormula(model, model.rxns, false);

% Create a map to store unique reactions by their formula
unique_rxns = containers.Map();
toRemove = []; % List of reactions to remove

for i = 1:length(model.rxns)
    formula = rxnFormula{i}; % Get reaction formula
    
    if isKey(unique_rxns, formula)
        % Get the index of an existing reaction
        existing_rxn = unique_rxns(formula);
        
        % Merge GPR rules only if they are different
        if ~strcmp(model.grRules{existing_rxn}, model.grRules{i})
            % Split GPR rules, remove duplicates, and format correctly
            uniqueGenes = unique([strsplit(model.grRules{existing_rxn}, ' or '), ...
                                  strsplit(model.grRules{i}, ' or ')]);
            model.grRules{existing_rxn} = strjoin(uniqueGenes, ' or ');
        end
        
        % Mark this reaction for removal
        toRemove = [toRemove, i];
        
    else
        % Store this reaction as unique
        unique_rxns(formula) = i;
    end
end
% **Delete all duplicate reactions in one step**
model = removeRxns(model, model.rxns(toRemove));
% **Display the optimized reaction count**
fprintf('Final reaction count: %d\n', length(model.rxns));