MALDI-TOF mass spectrometry of saliva samples as a prognostic tool for COVID-19

Figures & data

Figure 1. General workflow of the saliva sample preparation and MALDI-TOF analysis. (a) Method optimization. Initially, a range of different protein concentrations was evaluated to determine the optimal protein concentration. Subsequently, three MALDI matrices with better performances were evaluated and α-cyano-hydroxycinnamic acid (HCCA) was selected as the optimal one. (b) The sample cohort consisted of four groups: severe (ICU), mild (emergency room), moderate (department) and control. The COVID-19 diagnosis was confirmed by RT-qPCR of nasopharyngeal swabs. Saliva samples were obtained using a salivette and virus inactivation was performed in an BSL-3 laboratory. Direct spotting of saliva on the MALDI plate was performed and the proteomic profile was acquired using a MALDI-TOF MS instrument. The MALDI-TOF spectra were pre-processed and analyzed in R environment, where statistically significant protein peaks were used to train multiple machine learning algorithms for group classification (training phase). The model with the best performances was used to analyze independent samples and calculate the accuracy, sensitivity and specificity of the model (validation phase).

Table 1. Clinical characteristics of the 192 COVID-19 patients investigated

	MILD/MODERATE		SEVERE		Total		p^(a)
Patients characteristics	N	%	N	%	N	%
Sex
Male	59	53.6%	51	46.5%	110	100%	0.003*
Female	64	74.4%	22	25.6%	86	100%
Age
<40	30	69.8%	13	30.2%	43	100%	0.282
≥40	93	60.8%	60	39.2%	153	100%
Elderly
<60	88	65.7%	46	34.3%	134	100%	0.214
≥60	35	56.5%	27	43.5%	62	100%
Clinical features
Mechanical ventilation
None	59	96,7%	2	3,3%	61	100%	<0.001*
Non-invasive^a	64	60,4%	42	39,6%	106	100%
Invasive^b	0	0,0%	29	100,0%	29	100%
Vaccination status
Unvaccinated	106	61.6%	66	38.4%	172	100%	0.588
First dose only	10	66.7%	5	33.3%	15	100%
Fully vaccinated	7	77.8%	2	22.2%	9	100%
Outcome
Discharge	120	69.4%	53	30.6%	173	100%	<0.001*
Death	3	13.0%	20	87.0%	23	100%
Prevalence of symptoms according to disease severity
Fever	80/123	65.0%	41/73	56.2%	121/196	61.7%	0.216
Cough	80/123	65.0%	53/73	72.6%	133/196	67.9%	0.273
Headache	51/123	41.5%	21/73	28.8%	72/196	36.7%	0.075
Sore throat	28/123	22.8%	11/73	15.1%	39/196	19.9%	0.192
Muscle or body aches	53/123	43.1%	21/73	28.8%	74/196	37.8%	0.046*
Fatigue	52/123	42.3%	28/73	38.4%	80/196	40.8%	0.589
Congestion or runny nose	23/123	18.7%	9/73	12.3%	32/196	16.3%	0.243
Difficulty breathing	62/123	50.4%	52/73	71.2%	114/196	58.2%	0.004*
Loss of smell	19/123	15.4%	9/73	12.3%	28/196	14.3%	0.546
Loss of taste	21/123	17.1%	17/73	23.3%	38/196	19.4%	0.287
Nausea or vomiting	22/123	17.9%	13/73	17.8%	35/196	17.9%	0.989
Diarrhea	21/123	17.1%	8/73	11.0%	29/196	14.8%	0.244

ICU – Intensive Care Unit; *Statistical significance; (1) Pearson Chi-Square.

^aNasal cannula, facial mask ventilation or high flow nasal cannula.

^bInvasive mechanical ventilation.

Figure 2. (a) Mean spectra profile obtained from three different samples in triplicate at different concentrations (1:1, 1:2, 1:5, 1:10 and 1:100 sample to TFA 0.1% ratio). (b) Maximum intensity of the spectra for each concentration. (c) The number of peaks identified for each concentration. (d) Mean spectra profile of one sample in triplicate for each matrix tested. (e) Maximum intensity of the spectra for each matrix. (f) Number of peaks detected for each matrix. (g) PCA of the samples that were analyzed with 0.2 μg of proteins. (h) PCA of the samples using 1ul of saliva irrespective of the protein concentration. In Figure 2(g,h), MILD refers to MILD/MODERATE condition.

Figure 3. (a) MALDI-TOF MS mean spectrum of the saliva of the MILD/MODERATE group. (b) MALDI-TOF MS mean spectrum of the saliva of the SEVERE group. (c) MALDI-TOF MS mean spectrum of the saliva of the control group. (d) MALDI-TOF MS mean spectrum of the saliva of the infected group (MILD/MODERATE+SEVERE). (e) PCA for the control and infected groups. (f) The most ranked peaks after information gain filtering. (g) Spectra from the region of the highest ranked peak (1633). (h) Best ROC curves obtained from model training with control and infected samples. (i) Best PR curves obtained from model training with control and infected samples.

Figure 4. (a) PCA obtained through the analysis of the MILD/MODERATE and SEVERE conditions. (b) The 10 most relevant peaks were obtained through information gain filtering. (c) Comparison between MILD/MODERATE and SEVERE of the most discriminant peak. (d) Best ROC curves obtained from the model training. (e) Best PR curves obtained from the model training. (f) Mean accuracy, mean sensitivity, mean specificity and mean kappa were obtained in the model training step.