`_blood_correct_test_()` ignores `commu_score_col` and always corrects `Commu_Score`

[LeonYiFan]

Hi MEBOCOST developers,

I noticed that _blood_correct_test_() exposes a commu_score_col argument, and the official multisample demo calls it with commu_score_col='Norm_Commu_Score', but the current implementation appears to ignore this argument and always uses Commu_Score for the regression and correction.

Where

In src/mebocost/mebocost.py:

https://github.com/kaifuchenlab/MEBOCOST/blob/main/src/mebocost/mebocost.py#L1149-L1172

The argument is defined here:

commu_score_col = 'Commu_Score'

but the function body later hard-codes Commu_Score here:

r1, p1 = pearsonr(commu_res['blood_level'], commu_res['Commu_Score'])
sr1, sp1 = spearmanr(commu_res['blood_level'], commu_res['Commu_Score'])
model.fit(commu_res[['blood_level']], commu_res['Commu_Score'])
commu_res['Corrected_Commu_Score'] = commu_res['Commu_Score'] - commu_res['pred']

The plotting code also hard-codes Commu_Score:

sns.regplot(data=commu_res, x='blood_level', y='Commu_Score', ...)

Why this is confusing

The official multisample demo calls:

https://github.com/kaifuchenlab/MEBOCOST/blob/main/Demo_Multisample_mCCC_analysis.ipynb#L995-L999

corrected_commu_res = mebo_obj._blood_correct_test_(
    met_cont_file='data/mebocost_db/common/hmdb_blood_metabolite_concentration.tsv',
    commu_score_col='Norm_Commu_Score',
    ...
)

Based on this call, users would expect:

Corrected_Commu_Score = Norm_Commu_Score - predicted_blood_component

But the current function actually computes:

Corrected_Commu_Score = Commu_Score - predicted_blood_component

even when commu_score_col='Norm_Commu_Score' is passed.

Suggested fix

Use commu_score_col consistently inside _blood_correct_test_().

For example:

if commu_score_col not in commu_res.columns:
    raise KeyError(f"{commu_score_col} is not found in commu_res columns")

r1, p1 = pearsonr(commu_res['blood_level'], commu_res[commu_score_col])
sr1, sp1 = spearmanr(commu_res['blood_level'], commu_res[commu_score_col])

model = LinearRegression(fit_intercept=True)
model.fit(commu_res[['blood_level']], commu_res[commu_score_col])

commu_res['pred'] = model.predict(commu_res[['blood_level']])
commu_res['Corrected_Commu_Score'] = commu_res[commu_score_col] - commu_res['pred']

And for the first regression plot:

sns.regplot(
    data=commu_res,
    x='blood_level',
    y=commu_score_col,
    ci=False,
    ax=ax[0],
    scatter_kws={'alpha': .5}
)

Optionally, the y-axis label could also reflect the score column being corrected.

Expected behavior

When users pass:

commu_score_col='Norm_Commu_Score'

the returned Corrected_Commu_Score should be based on Norm_Commu_Score, not Commu_Score.

This would make the implementation consistent with the official demo and with the function argument.

Thank you!