Dataset preprocessing parameters

Table 1 sets out all the running parameters needed to fully configure the preprocessing stage. Due to the different data distributions, the baseline removal, normalization, peak detection SNR and correlation threshold $\rho$ were the same for the four datasets, while the other parameters needed to be adapted individually. Notice that all the preprocessing steps are performed for each external fold of the DAW (see Section V of the paper).

Table 1: Running parameters configured for the preprocessing task. Note that the first four parameters are shared across all the datasets, while the last three need to be adapted individually. (*)The formulation represents a 300-position binary array containing ones except for each end position whose value is zero.

Top-hat structural mask	[0; ones(298,1)*; 0]
Window size for normalization scanning	200
Minimum SNR on peak detection	3
Threshold $\rho$ to agglomerate peakbins	0.80
	OVA	TOX	HCC	DGB
Smoothing wavelet threshold	5	6	6	6
Power of 2 for wavelet denoising	3	7	10	10
Minimum intensity $T$ on peak detection	5	20	4000	1500