Even though the COVID-19 pandemic is relatively resolving, it is essential to find techniques to monitor the disease in case of its reactivation or virus mutation. Therefore, in this work, we analyzed the feasibility of using 2T2D-COS employing FTIR spectra to monitor the progression of COVID-19.
The symptoms of COVID-19, the comorbidities, the blood type related to the disease, and various biochemical parameters in vaccinated and non-vaccinated patients have been discussed extensively in other articles comparing populations with considerable numbers of patients[25, 28, 29].
Likewise, we have correlated the biochemical and clinical parameters with FTIR spectra; however, we found interesting data in the hematic biometry test and immunological response when grouped by age to analyze the 2T2D-COS.
In this sense, neutrophils are the first responder of immune cells, followed by lymphocytes, and help the body fight infection and other diseases. The usual range of neutrophils in a healthy adult is between 1,700 and 6,500 per microliter of blood. The essential infection-fighting white blood cell (WBC) is the neutrophil. In this study, a high value of neutrophils was observed in males with COVID-19 of 25, 30, 35, 50, and 75 years. For the female population, its value was high for 35, 50, 65, 70, and 75 age.
Lymphocytes have a slower but significant response to inflammation and tend to arrive after neutrophils as part of the adaptive immune response. In adults, the normal range of lymphocytes is between 1,000 and 3,200 per microliter of blood. If the lymphocyte count is high, the test result might be evidence of infection (bacterial or viral). In our patients, this value falls within the permissible limit for both the male and female populations.
In the same way, the NLR is a sensitive indicator of infection, inflammation, and sepsis, validated in numerous studies. Furthermore, clinical research confirmed the sensitivity of NLR for diagnosis/stratification of systemic infection, sepsis, and bacteremia, as well as its substantial predictive and prognostic value. Therefore, it is recommended that the NLR should be investigated daily and follow-up its absolute values and dynamic course in an acute disease or critical illness[30]. Regarding this, we found that this indicator increased in all age groups.
Moreover, the leukocytes are WBC that play a critical role in immunity. Therefore, a high WBC count usually means a bacterial infection reaction. The specific number of leukocytes in the blood is 3,800 to 9,800 per microliter. Thus, biochemical studies allow us to demonstrate the systemic infection by SARS-CoV-2 and some effects on the organism. Therefore, to gain insight into the immune function, we studied 2T2D-COS using FTIR spectra to search for a mechanism correlating the antibody response in saliva.
2T2D-COS spectral analysis of male populations in the 3000 − 2000 cm−1 region.
Asynchronous 2T2D cross peaks +(3074, 2582), +(3074, 2246), +(2937,2582), and +(2937,2246) (Fig. 2a.(e)) show the abundant presence of amide B of proteins α amylase and albumin than 2582 cm− 1 of SH thiol groups and 2546 of N = C = O stretch of isocyanide in M25 COVID-19 saliva samples. This lower abundance of SH thiol shows the strong interaction of hydrogen bond influence with SARS-CoV-2 protein as a defense mechanism. The asynchronous spectra − (3083,2921) (Fig. 2a(g) and Fig. 2b(e)) of male M35 and M45 shows higher amount of υas CH2 lipids and fatty acids in control compared to COVID-19 samples. This indicates lipid synthesis to overcome energy in the defense mechanism of SARS-CoV-2. Significantly increased absorbances were also reported by Kazmer, S.T. et al. (31) in SARS-CoV-2 infection induced in both cell and mouse models using an FTIR study. The asynchronous spectra of 55, 60, 70, and 75 (Figs. 2b(f,h) and 2c(e,f)) -(3080, 2878) indicates υas CH2 lipids fatty acids decreases due to oxidation of lipids in SARS-CoV-2 saliva samples. This shows υas CH2 lipids and fatty acids decrease, resulting in lipid oxidation resisting the SARS-CoV-2. Duś-Ilnicka I.[32] reported that absorption bands of CH– and CH2-groups characterize changes in the structure of lipids and show a more intense course of lipid peroxidation.
2T2D-COS spectral analysis of female populations in the 3000 − 2000 cm−1 region.
The asynchronous spectra of F25 and F30 (Fig. 2a(m & n)) cross peak –(3085, 2924) show a decrease in the intensity of lipids and fatty acids. It indicates oxidation of lipids due to SARS-CoV-2. A minute differentiation of specific bands is the advantage of 2T2D correlation, which is impossible using conventional spectra. Xiang et al. have also noticed a cluster of –CH2 and –CH3 stretching vibrations in the spectral range of 2800–3100 cm− 1 (mainly attributed to lipids). The cross peaks +(2928,2058) and +(3075, 2058) show a decrease in thiocyanate ~ 2058 cm− 1 in COVID-19 samples. It shows the role of SCN− 1 ions in the defense mechanism due to SARS-CoV-2.
2T2D-COS spectral analysis of male populations in the 1800 − 900 cm−1 region.
The auto peak ~ 1643 cm− 1 indicates amide I of proteins, an identified biomarker responsible for the course of COVID-19[33]. The asynchronous spectra –(1657,1523) M25 (Fig. 3a(e) shows the presence of high IgG in control compared to COVID-19 samples. The asynchronous spectra of M30 populations (Fig. 3a(f)) show that IgG and IgM are more in control than COVID-19 populations. Zhao J. et al.[34] recorded viral infections attributed to Ig protein marker in positive serum spectra induced by hepatitis B and C virus. Our study using 2T2D shows a clear distinction, as evidenced by the change in IgG and IgM due to the SARS-CoV-2 compared to the control. The male 35, 40, 45, 55, and 75 COVID-19 populations (Figs. 3a(g,h), 3b(e,g) and 3c(f)) show cross peaks –(1656,1523), which signifies the presence of less abundance or weak IgG. It indicates SARS-COV-2 envelop due to spike receptor binding domain (RBD). Thus, IgG is susceptible to SARS-CoV-2 in M35, 40, 45, 55, and 75. In addition, less abundance of amide I proteins shows the high binding affinity of amide I proteins for this population group due to SARS-CoV-2.
The -ve cross-peaks observed for M50 -(1650,1397), -(1648,1402) show a decrease in IgM level (~ 1397–1402) and signifies the role of the immune system’s acts as a defense against the virus[33]. Thus, 2T2D resolves and identify the sensitive, weak peak to the SARS-CoV-2.
2T2D-COS spectral analysis of female populations in the 1800-900 cm-1 region
F30 population +(1643,1109), +(1532,1074) (Fig. 3a(j)) show higher presence IgG > IgM. Thus, IgG response displays a major contributor to the protective effect against COVID-19 associated with other factors such as sex or age [35, 36]. Isho et al.[37] reported maximum IgA levels (~ 74.1% and ~ 84.2%), while IgM (66.2% and ~ 75.1%) levels were less compared to IgA. Thus, IgG responses in saliva may serve as a measure of systemic immunity to SARS-CoV-2. Our study confirms that saliva IgG antibodies to SARS-CoV-2 are maintained in most COVID-19 patients. F45 (Fig. 3b(m)) shows the induction protein aggregating through spike proteins SARS-CoV-2 influences amide I proteins[38]. The protein bands, on average, appear more intense in the positive samples than the negative ones. Experiments from FTIR spectra of SARS-CoV-2 infection in the controlled cell and mouse model infection show increased absorbance at the amide I band (1700–1600 cm− 1) in the active SARS-CoV-2 sample, thus providing proof-of-specificity[31]. For F55, 60, 65, 70, and 75 years (Fig. 3b(o,p) and Fig. 3c(k-l)) shows the presence of two -ve cross peaks –(1654,1525) and − (1653,1401) evidencing the higher value of IgG and IgM in control samples compared to the COVID-19 population. In another way, less abundance of IgG and IgM is observed for the female COVID-19 samples of 55, 60, 65, 70, and 75 years. This clearly shows that SARS-CoV-2 influences immunoglobulin IgG/IgM proteins. Thus, 2T2D shows minute spectral changes observed compared to control for the respective age of samples.
Performance of measures male population
ROC analysis of FTIR spectra helps to diagnose biological samples with high sensitivity and specificity[19, 39, 40]. It is a statistically valid method for biomarker evaluation. Wood et al.[38] 2021, reported FTIR study of saliva samples shows excellent results with an AUC of 0.90. Zozan Guleken et al.[33] studied using FTIR with distinct spectral differences between standard control and COVID-19 with the area under the receiver operating characteristic curve (AUC) of 0.95. Our findings from ROC have high values ranging from (0.81–0.96) with our results. Caixeta et al.[41] used ROC analysis from the FTIR spectra for diabetic with non-diabetic rats where the sensitivity and specificity were 100% and 93.3%, respectively, with an AUC of 0.98. ROC curve analysis is widely considered to be the most objective and statistically valid method for biomarker performance evaluation[41, 42]. ROC analysis gives information about the overall efficiency of a diagnostic study for the different groups of male and female populations. The measured area under the curve (AUC) signifies the high reliability of the test. Figure 4a and 4b shows the ROC curve for different age (25, 30, 35, 40, 45, 50, 55, 60, 65, 70, and 75 years) groups of the male and female population. In the male population (Table 1) the measured AUC values range from 0.81 to 0.96, with minimum values observed for M30 and higher values for M60. The various computed parameters are shown in Table 3. The constructed confusion matrix for the respective age groups of male populations is shown in Table 3. The sensitivity value was more significant of 0.90% for all the samples except for M30, 35, 45, and 70 years (Table 1). The F1 score gives the overall performance of measures. A high value of above 90% was observed for M25, 50, 55, 60, 65, and 75. The F1 score for the remaining samples (M30, 35, 40, 45) lies in the 74–90% range. The G1 score measured overall classification performance irrespective of change in sample size. Its values are high, above 90% for the sample for M25, 50, 55, 60, 65, and 75, and 74–89% for the rest of the samples. This signifies high stability of biochemical compositions in the saliva of higher age groups studied except for M25 years.
Table 1
The various computed performance of measures of different male population group.
Age of Sample
|
Sensitivity %
|
Specificity %
|
Area Under Curve (AUC)
|
F1 score %
|
G1 score %
|
PPV%
|
NPV%
|
LR +
|
LR-
|
MCC%
|
Accuracy %
|
25
|
0.95
|
0.86
|
0.94
|
90.91
|
90.98
|
50.00
|
100.00
|
7.57
|
0.06
|
0.66
|
88.24
|
30
|
0.79
|
0.70
|
0.81
|
74.52
|
74.67
|
35.29
|
92.86
|
2.65
|
0.29
|
0.36
|
71.11
|
35
|
0.88
|
0.71
|
0.81
|
78.61
|
79.08
|
64.29
|
92.86
|
3.03
|
0.17
|
0.60
|
78.57
|
40
|
0.99
|
0.81
|
0.93
|
89.24
|
89.72
|
40.00
|
100.00
|
5.21
|
0.01
|
0.57
|
82.35
|
45
|
0.83
|
0.86
|
0.89
|
84.73
|
84.74
|
63.16
|
95.92
|
6.04
|
0.19
|
0.66
|
86.76
|
50
|
0.93
|
0.90
|
0.96
|
91.32
|
91.34
|
60.00
|
97.73
|
9.03
|
0.08
|
0.67
|
89.38
|
55
|
0.91
|
0.91
|
0.92
|
90.95
|
90.95
|
76.47
|
96.00
|
9.71
|
0.10
|
0.76
|
89.91
|
60
|
0.93
|
0.94
|
0.96
|
93.35
|
93.35
|
88.10
|
95.71
|
14.34
|
0.07
|
0.85
|
92.86
|
65
|
0.91
|
0.90
|
0.92
|
90.45
|
90.45
|
85.71
|
94.59
|
9.43
|
0.11
|
0.81
|
90.77
|
70
|
0.86
|
0.80
|
0.89
|
82.99
|
83.05
|
57.14
|
94.44
|
4.34
|
0.17
|
0.58
|
81.71
|
75
|
0.99
|
0.89
|
0.94
|
93.70
|
93.83
|
85.71
|
100.00
|
9.06
|
0.01
|
0.88
|
93.75
|
Mathew correlation coefficient MCC has advantages that measure the binary classification of unbalanced data sets. Its values range between (-1, + 1), where − 1 is related to perfect misclassification and + 1 to perfect classification (23). The MCC value is higher, which ranges from 0.76–0.88 for the M55, 60-, 65-, and 75-year samples. This shows the best classification of COVID + ve and -ve saliva samples constructed from the confusion matrix. This was further justified by the 81–93% accuracy obtained for the respective samples. The most negligible MCC value of 0.36 was obtained for M30 samples, which may be due to the decrease in the size of the samples. The high LR + ad low LR- value obtained for M50, 55, 60, 65, and 75 gives a good perception of the classification of the samples in discriminating SARS-CoV-2 in the saliva samples. For M30 and 35, reasonable accuracy, F1 score, G1 score, and AUC value are obtained. But LR + and LR- and MCC values do not agree with the quality performance of the samples studied. This may be improved with a significant increase in samples. Thus, in addition to F1, G1 showed a higher performance precision was assessed by LR + and LR- and MCC values.
Performance of measures of the female population
The ROC analysis was performed for female populations of different ages (Fig. 4b(a-k)). The results of ROC analysis show a high value of AUC lies in the range of 0.81–0.96 with good sensitivity (0.81-.99) and specificity (0.74–0.92) in the female samples studied (Table 2). Nogueira et al.[43] reported that the AUC was 0.82 showing satisfactory performance for real-time COVID-19 detection. The accuracy of test performance has a good value of 90% and above for the samples F25, 40, 50, 55, and 65. The remaining samples have an accuracy range of 71–85%. The computed F1 and G1 scores have a high value in the 73–91% range. A high PPV value above 90% was observed for F40, 45, 55, and 70 years of saliva samples. The least value was obtained for 66.67% and 57.14% for F25 and 35, respectively. The NPV value of 90% and above were obtained for F40, 50, 55, 60, and 65. This shows that predicated + ve and -ve COVID samples have good accuracy. The high LR + and low LR- values were obtained for the samples studied. This MCC value is the next parameter in assessing the distinction of samples of COVID + ve and-ve from unbalanced data, even though it has a good accuracy value.
Table 2
The various computed performance of measures of different female population group.
Age of Sample
|
Sensitivity %
|
Specificity %
|
Area Under Curve (AUC)
|
F1 score %
|
G1 score %
|
PPV%
|
NPV%
|
LR +
|
LR-
|
MCC%
|
Accuracy %
|
25
|
0.90
|
0.89
|
0.92
|
89.70
|
89.70
|
66.67
|
100.00
|
8.28
|
0.11
|
0.77
|
90.91
|
30
|
0.86
|
0.74
|
0.81
|
79.35
|
79.57
|
88.24
|
77.78
|
3.27
|
0.19
|
0.66
|
84.62
|
35
|
0.80
|
0.68
|
0.80
|
73.70
|
73.93
|
57.14
|
85.71
|
2.53
|
0.29
|
0.45
|
71.43
|
40
|
0.99
|
0.84
|
0.95
|
90.99
|
91.27
|
94.12
|
92.86
|
6.33
|
0.02
|
0.85
|
93.75
|
45
|
0.85
|
0.89
|
0.92
|
87.00
|
87.03
|
93.10
|
76.19
|
7.80
|
0.17
|
0.71
|
86.00
|
50
|
0.91
|
0.92
|
0.95
|
91.00
|
91.00
|
88.24
|
92.31
|
10.65
|
0.10
|
0.81
|
90.52
|
55
|
0.99
|
0.92
|
0.96
|
95.46
|
95.53
|
91.67
|
98.04
|
12.26
|
0.01
|
0.90
|
94.95
|
60
|
0.86
|
0.76
|
0.87
|
81.09
|
81.25
|
68.89
|
90.20
|
3.66
|
0.18
|
0.61
|
80.21
|
65
|
0.95
|
0.84
|
0.94
|
89.33
|
89.49
|
89.66
|
90.91
|
6.05
|
0.06
|
0.79
|
90.11
|
70
|
0.86
|
0.90
|
0.96
|
87.74
|
87.77
|
95.00
|
75.86
|
8.64
|
0.16
|
0.73
|
86.96
|
75
|
0.80
|
0.91
|
0.94
|
85.14
|
85.29
|
88.46
|
81.82
|
8.54
|
0.22
|
0.70
|
84.75
|
Table 3
Confusion matrix representing true positive (TP), false positive (FP), true negative (TN), and false negative (FN) of male and female population 25, 30, 35, 40, 45, 50, 55, 60, 65, 70, and 75 respectively.
Predict
|
Predict
|
Actual
|
25M
|
COVID-19
|
Control
|
Actual
|
25F
|
COVID-19
|
Control
|
COVID-19
|
12(TN)
|
3(FP)
|
COVID-19
|
13(TN)
|
2(FP)
|
Control
|
0(FN)
|
2(TP)
|
Control
|
0(FN)
|
2(TP)
|
Predict
|
Predict
|
Actual
|
30M
|
COVID-19
|
Control
|
Actual
|
30F
|
COVID-19
|
Control
|
COVID-19
|
26(TN)
|
11(FP)
|
COVID-19
|
15(TN)
|
2(FP)
|
Control
|
2(FN)
|
6(TP)
|
Control
|
2(FN)
|
7(TP)
|
Predict
|
Predict
|
Actual
|
35M
|
COVID-19
|
Control
|
Actual
|
35F
|
COVID-19
|
Control
|
COVID-19
|
13(TN)
|
5(TP)
|
COVID-19
|
13(TN)
|
5(TP)
|
Control
|
1(FN)
|
9(TP)
|
Control
|
2(FN)
|
8(TP)
|
Predict
|
Predict
|
Actual
|
40M
|
COVID-19
|
Control
|
Actual
|
40F
|
COVID-19
|
Control
|
COVID-19
|
36(TN)
|
9(FP)
|
COVID-19
|
13(TN)
|
2(FP)
|
Control
|
0(FN)
|
6(TP)
|
Control
|
1(FN)
|
32(TP)
|
Predict
|
Predict
|
Actual
|
45M
|
COVID-19
|
Control
|
Actual
|
45F
|
COVID-19
|
Control
|
COVID-19
|
47(TN)
|
7(FP)
|
COVID-19
|
6(TN)
|
2(FP)
|
Control
|
2(FN)
|
12(TP)
|
Control
|
5(FN)
|
27(TP)
|
Predict
|
Predict
|
Actual
|
50M
|
COVID-19
|
Control
|
Actual
|
50F
|
COVID-19
|
Control
|
COVID-19
|
86(TN)
|
10(TP)
|
COVID-19
|
60(TN)
|
6(FP)
|
Control
|
2(FN)
|
15(TP)
|
Control
|
5(FN)
|
45(TP)
|
Predict
|
Predict
|
Actual
|
55M
|
COVID-19
|
Control
|
Actual
|
55F
|
COVID-19
|
Control
|
COVID-19
|
72(TN)
|
8(FP)
|
COVID-19
|
50(TN)
|
4(FP)
|
Control
|
3(FN)
|
26(TP)
|
Control
|
1(FN)
|
44(TP)
|
Predict
|
Predict
|
Actual
|
60M
|
COVID-19
|
Control
|
Actual
|
60F
|
COVID-19
|
Control
|
COVID-19
|
67(TN)
|
5(FP)
|
COVID-19
|
46(TN)
|
14(FP)
|
Control
|
3(FN)
|
37(TP)
|
Control
|
5(FN)
|
31(TP)
|
Predict
|
Predict
|
Actual
|
65M
|
COVID-19
|
Control
|
Actual
|
65F
|
COVID-19
|
Control
|
COVID-19
|
35(TN)
|
4(FP)
|
COVID-19
|
30(TN)
|
6(FP)
|
Control
|
2(FN)
|
24(TP)
|
Control
|
3(FN)
|
52(TP)
|
Predict
|
Predict
|
Actual
|
70M
|
COVID-19
|
Control
|
Actual
|
70F
|
COVID-19
|
Control
|
COVID-19
|
51(TN)
|
12(TP)
|
COVID-19
|
22(TN)
|
2(TP)
|
Control
|
3(FN)
|
16(TP)
|
Control
|
7(FN)
|
38(TP)
|
Predict
|
Predict
|
Actual
|
75M
|
COVID-19
|
Control
|
Actual
|
75F
|
COVID-19
|
Control
|
COVID-19
|
18(TN)
|
2(FP)
|
COVID-19
|
27(TN)
|
3(FP)
|
Control
|
0(FN)
|
12(TP)
|
Control
|
6(FN)
|
23(TP)
|