Multiplex plasma protein assays as a diagnostic tool for lung cancer

Abstract

Lack of the established noninvasive diagnostic biomarkers causes delay in diagnosis of lung cancer (LC). The aim of this study was to explore the association between inflammatory and cancer-associated plasma proteins and LC and thereby discover potential biomarkers. Patients referred for suspected LC and later diagnosed with primary LC, other cancers, or no cancer (NC) were included in this study. Demographic information and plasma samples were collected, and diagnostic information was later retrieved from medical records. Relative quantification of 92 plasma proteins was carried out using the Olink Immuno-Onc-I panel. Association between expression levels of panel of proteins with different diagnoses was assessed using generalized linear model (GLM) with the binomial family and a logit-link function, considering confounder effects of age, gender, smoking, and pulmonary diseases. The analysis showed that the combination of five plasma proteins (CD83, GZMA, GZMB, CD8A, and MMP12) has higher diagnostic performance for primary LC in both early and advanced stages compared with NC. This panel demonstrated lower diagnostic performance for other cancer types. Moreover, inclusion of four proteins (GAL9, PDCD1, CD4, and HO1) to the aforementioned panel significantly increased the diagnostic performance for primary LC in advanced stage as well as for other cancers. Consequently, the collective expression profiles of select plasma proteins, especially when analyzed in conjunction, might have the potential to distinguish individuals with LC from NC. This suggests their utility as predictive biomarkers for identification of LC patients. The synergistic application of these proteins as biomarkers could pave the way for the development of diagnostic tools for early-stage LC detection.

Keywords: confounder correction; diagnostic biomarker; liquid biopsy; lung cancer; machine learning; plasma proteomics; screening.

FIGURE 1

Clinical characteristics (LC histology, gender, mutation, stage of cancer, smoking status, and concurrent pulmonary disease status) on clustered plasma protein expression. ADC, adenocarcinoma; LC, lung cancer; SCC, squamous cell carcinoma; SCLC, small cell lung cancer.

FIGURE 2

Expression of proteins of panel 1 (CD83, GZMA, GZMB, CD8A, and MMP12) in early an advanced stage of different types of LC patients. NPX values on the y‐axis and different groups of patients based on LC histology in the x‐axis. (A) Half violin and box plot for panel 1 proteins: The left‐side violin plot shows the density of protein expression, while the right‐side box plot displays the data distribution across the first (Q1), second (Q2), and third (Q3) quartiles, including outliers. The brown dot represents the mean for each disease group. (B) The mean expressions of individual proteins in different disease groups are shown compared with the no cancer group. ADC, adenocarcinoma; LC, lung cancer; nscLC, non‐small cell lung cancer SCC, squamous cell carcinoma; scLC, small cell lung cancer.

FIGURE 3

Classification (AUC) and diagnostic (Youden index) performance for the panel of proteins including their sensitivity and specificity for panels 1, 2, 3, 6, 7, 8, and 9 against the no cancer patient group. (A) Classification performance by AUC. (B) Diagnostic performance by Youden index. (C) Sensitivity. (D) Specificity. For all LC subtypes in different stratified analysis panels 1, 7, 8, and 9 showed higher values in all matric compared to panels 2, 3 and 6. Only for other cancer group, panel 6 showed higher performance in AUC and Youden index. ADC, adenocarcinoma; AUC, area under the receiver‐operating characteristic curve; LC, lung cancer; NSCLC, non‐small cell lung cancer; PLC, Primary lung cancer; SCC, squamous cell carcinoma; SCLC, small cell lung cancer.

FIGURE 4

(A) No grouping tendency was observed in the expression of proteins (panels 1, 3, 6, and 8) in different cancer groups (no cancer, early‐stage LC, advanced‐stage LC, and other cancer) of patients with different characteristics (histology of LC tumor, gender, status of pulmonary diseases, and smoking). (B) Correlation (Pearson) in the expression of proteins (in panels 1, 2, and 6). The values are the strength of correlation, and insignificant correlations (p ≥ 0.05 in Pearson's coefficient) are left blank. (C–E) Area under the receiver‐operating characteristic curve (AUROC) values and Youden index of panels (panels 1, 6, and 8). Also, single proteins in the panels that are significant (p ≤ 0.05) between different patient groups (early, advanced, and other cancer) compared with no cancer are demonstrated. ADC, adenocarcinoma; LC, lung cancer; SCC, squamous cell carcinoma; SCLC, small cell lung cancer.