**2. Results**

Plasma and urine samples were collected from 15 patients with diabetic nephropathy (group D), from 14 patients with glomerulonephritis (group G), from 5 patients with hypertensive nephropathy (group H), and from 14 healthy volunteers (group N). Table 1 summarizes the principal characteristics of the CKD patients.


**Table 1.** Principal characteristics of the chronic kidney diseases (CKD) patients.

Abbreviations: SD, standard deviation; eGFR, estimated glomerular filtration rate.

The plasma samples were depleted from albumin and immunoglobulin, the urine samples were concentrated by filter columns. All protein samples were prepared for mass spectrometry in duplicates and each duplicate was analyzed twice. These technical replicates were marked as A and B. The total set of plasma probes contained 190 samples (group G—56, group D—58, group H—20, and group N—56). The number of urine samples derived from patients was 97 (group G—40, group D—41, and group H—16); urine from healthy people was not taken due to low normal protein concentrations. The samples were not divided into fractions to limit the resulted data set for the convenience of bioinformatic processing. Two sets of quantitative proteomics data from blood and urine samples (Table S1) were obtained. No specific differences on proteome profiles between different patients groups expressed in the distribution of individual proteins were found by conventional statistical methods. Thus, machine learning algorithms, taking into account the overall contribution of the proteomic composition of the samples, were applied. The data were independently tested in two ways. At first, we distinguished the total CKD patients from the healthy individuals, and then we differentiated three groups of patients from each other without comparison with healthy control.
