Question regarding the input data for analysis: Aligned (Warped) vs. Preprocessed Spectra Matrix?

[conbercept]

First of all, thank you for your excellent work.

I am currently studying your paper and code with the intention of applying your workflow to data generated by other machines (specifically Vitek MS). However, I have encountered some confusion regarding the exact input data used for the downstream analysis and would appreciate your clarification.

1. Paper Description In the methodology section of your paper, standard preprocessing steps are described as follows:

"The following preprocessing steps are performed using the R package MaldiQuant58 v1.19: (1) the measured intensity is transformed with a square-root method to stabilize the variance, (2) smoothing using the Savitzky–Golay algorithm with half-window-size 10 is applied, (3) an estimate of the baseline is removed in 20 iterations of the SNIP algorithm, (4) the intensity is calibrated using the total ion current (TIC), and (5) the spectra are trimmed to values in a 2,000–20,000 Da range."

2. Observation in Code However, upon reviewing your scripts, I noticed a process for Species-specific Reference Peak Construction, which includes:

Peak Detection: MAD-based peak detection with an SNR threshold of 3.

Frequency Analysis: Minimum frequency threshold of 90% across species samples.

Tolerance Matching: 0.004 Da tolerance for peak alignment.

Reference Generation: Creation of consensus peak sets for major pathogens.

3. My Question Could you please clarify which dataset was used for the final data analysis/classification tasks?

Was it the aligned/warped peak matrix (e.g., DRIAMS-A_2015_warped.r) that has been aligned to the reference peaks?

Or was it the standard preprocessed matrix (e.g., DRIAMS-A_2015_preprocessed.r) without the species-specific peak alignment?

Understanding this step is crucial for me to correctly adapt the pipeline for Vitek MS data.

Thank you very much for your time and assistance!