The Role of Genomics and Proteomics in Lung Cancer Early Detection and Treatment

Simple Summary Biomarkers for the prediction and efficacy of therapies are an urgent necessity for lung cancer patients. In this article, we summarize genomic and proteomic biomarkers utilized for the early detection and treatment of lung cancer, with a focus on immune checkpoint and PI3K pathways. Abstract Lung cancer is the leading cause of cancer-related death worldwide, with non-small-cell lung cancer (NSCLC) being the primary type. Unfortunately, it is often diagnosed at advanced stages, when therapy leaves patients with a dismal prognosis. Despite the advances in genomics and proteomics in the past decade, leading to progress in developing tools for early diagnosis, targeted therapies have shown promising results; however, the 5-year survival of NSCLC patients is only about 15%. Low-dose computed tomography or chest X-ray are the main types of screening tools. Lung cancer patients without specific, actionable mutations are currently treated with conventional therapies, such as platinum-based chemotherapy; however, resistances and relapses often occur in these patients. More noninvasive, inexpensive, and safer diagnostic methods based on novel biomarkers for NSCLC are of paramount importance. In the current review, we summarize genomic and proteomic biomarkers utilized for the early detection and treatment of NSCLC. We further discuss future opportunities to improve biomarkers for early detection and the effective treatment of NSCLC.


Introduction
Worldwide, the second most common type of cancer diagnosed in men and women is lung cancer, which is the most common cause of cancer-related deaths. There are two main types of lung cancers: non-small-cell lung cancer (NSCLC) and small-cell lung cancer (SCLC). Adenocarcinoma and squamous cell carcinoma are the main histological subtypes of NSCLC. Prostate, colon, and breast cancers have screening tools for early detection; however, there are generally less accurate early detection biomarkers for lung cancer [1]. GLOBOCAN estimated approximately 2.2 million new lung cancer cases and 1,796,144 deaths [2].
The five-year survival rate in stage I NSCLC patients is 70-90%. However, over 75% of lung cancer patients are diagnosed at advanced or metastatic stages when the five-year survival rates are significantly lower [2,3]. Its diagnosis at advanced stages is associated with significantly lower five-year survival.

Genomic and Epigenomes
2.1.1. Methylation DNA methylation (hypermethylation or hypomethylation) is an epigenetic modification that plays a pivotal role in silencing gene transcription and is crucial in many biological processes, including embryonic development, X-chromosome inactivation, genomic imprinting, chromatin structure, normal growth, and cellular proliferation. Abnormal DNA methylation has been observed in tumor initiation and proliferation of all forms of cancers [22]. The epigenetic pathway is potentially reversible and is involved in tumorigenesis by three main mechanisms: DNA hypermethylation, global hypomethylation of the genome, and histone modifications.
DNA methylation can be assessed by different techniques, such as digital PCR (dPCR), methylation-specific PCR (MSP), methyl-CpG-binding domain (MBD), real-time quantitative MSP, quantitative methylation-specific PCR (qMSP), multiplex nested methylation-specific PCR, and whole-genome bisulfite sequencing (WGBS) [23]. Aberrant DNA methylation is catalyzed by DNA methyltransferases (DNMTs) through the promoter methylation of tumor suppressor genes. It is associated with many types of cancers, including NSCLC, autoimmune diseases, diabetes, and multiple sclerosis [24][25][26]. Hypermethylation of DNA can silence tumor suppressor genes, and hypomethylation of DNA can lead to transcriptional activation proto-oncogenes. Methylation patterns have been associated with different cancer aspects, such as disease stage, survival, and response to 2.1. Genomic and Epigenomes 2.1.1. Methylation DNA methylation (hypermethylation or hypomethylation) is an epigenetic modification that plays a pivotal role in silencing gene transcription and is crucial in many biological processes, including embryonic development, X-chromosome inactivation, genomic imprinting, chromatin structure, normal growth, and cellular proliferation. Abnormal DNA methylation has been observed in tumor initiation and proliferation of all forms of cancers [22]. The epigenetic pathway is potentially reversible and is involved in tumorigenesis by three main mechanisms: DNA hypermethylation, global hypomethylation of the genome, and histone modifications.
DNA methylation can be assessed by different techniques, such as digital PCR (dPCR), methylation-specific PCR (MSP), methyl-CpG-binding domain (MBD), real-time quantitative MSP, quantitative methylation-specific PCR (qMSP), multiplex nested methylationspecific PCR, and whole-genome bisulfite sequencing (WGBS) [23]. Aberrant DNA methylation is catalyzed by DNA methyltransferases (DNMTs) through the promoter methylation of tumor suppressor genes. It is associated with many types of cancers, including NSCLC, autoimmune diseases, diabetes, and multiple sclerosis [24][25][26]. Hypermethylation of DNA can silence tumor suppressor genes, and hypomethylation of DNA can lead to transcriptional activation proto-oncogenes. Methylation patterns have been associated with different cancer aspects, such as disease stage, survival, and response to treatment. DNA methylation could predict the response to platinum-based chemotherapy in lung cancer patients [27,28]. tion, differentiation, and survival [44]. In cancer, miRNAs may act as either tumor suppressors or oncogenes [45,46]. Profiling miRNA in plasma, serum, and sputum are promising noninvasive biomarkers for the early detection of NSCLC (Table 2). Numerous studies have reported the great potential of miRNAs in lung cancer diagnosis. A recent systematic review and meta-analysis by Wang et al. declared that miRNA-21 has potential clinical value in the diagnosis and prognosis of lung cancer [47]. Lu et al. investigated 723 human microRNAs in 106 plasmas in healthy individuals and patients with NSCLC or SCLC. They identified the diagnostic features of six miRNAs (miR-17, miR-190b, miR-19a, miR-19b, miR-26b, and miR-375) that have value for discriminating lung cancer patients from healthy individuals [48]. A panel of three miRNAs (miR-125a-5p, miR-25, and miR-126) in the study of Wang et al. distinguished early stage lung cancer patients from control subjects [49]. Pan et al. studied the expression of miR-33a-5p and miR-128-3p lung cancer tissues and cell lines. They found that the expression of these miRNAs was downregulated in tissues, cell lines, and the whole blood of early stage lung cancer patients [50]. A multicenter study analyzed bloodborne RNA signatures in 3102 patients and found that a 15-miRNA signature might help distinguish patients diagnosed with lung cancer from cancer-free individuals [51]. Khandelwal et al. explored the role of miR-590-5p in NSCLC patients in plasma samples, and they found that miRNA functions as a tumor suppressor in NSCLC and downregulated in NSCLC patients compared to healthy controls [52].

Proteomic Early Detection
Proteomics is a powerful approach to studying numerous proteins that generates information about biomarker identification, molecular interactions, and signaling pathways [60]. Mass spectrometry (MS) is an analytical tool in biomarker discovery that can be used to measure the mass-to-charge ratio of particles. MS-based proteomics has been applied to biomarker discovery for several decades. In oncology, MS approaches provided a decisive result for the early detection of breast, colorectal, prostate, and ovarian cancer [61][62][63][64]. Proteomic biomarkers could be used from several sources, including plasma, serum, tears, sputum, saliva, and urine [65][66][67]. We can also find proteomic biomarkers in exhaled breath condensate, bronchoalveolar lavage fluid, and pleural effusion for lung cancer [68] (Table 3). Zhang et al. profiled urine proteome in several cancer types and healthy control. In this study, five biomarkers (FTL: ferritin light chain; MAPK1IP1L: mitogen-activated protein kinase 1 interacting protein 1-like; FGB: fibrinogen beta chain; RAB33B: RAB33B, member RAS oncogene family; RAB15: RAB15, member RAS oncogene family) could distinguish lung cancer patients from healthy individuals [69].

Genomics Biomarkers
In the United States, the National Cancer Institute (NCI) and National Human Genome Research Institute launched a genomic cancer program, The Cancer Genome Atlas (TCGA), in 2006. This project characterized 20,000 primary cancer and matched standard samples of 33 major cancer types in the genomic, epigenomic, transcriptomics, clinical, and proteomics information level [77]. Over several years, TCGA data has been available in several databases, such as cBioPortal for Cancer Genomics and The Cancer Proteome Atlas Portal (TCPA) [78,79]. The Catalogue of Somatic Mutations in Cancer (COSMIC) and the International Cancer Genome Consortium (ICGC) provide a comprehensive genomic data visualization, integration, and analysis source [80,81]. Integrative analysis from the Pan-Cancer Analysis of Whole Genomes (PCAWG) Consortium of the International Cancer Genome Consortium (ICGC) and TCGA revealed that cancer genomes contained four to five driver mutations [82]. Recent results from the PCAWG Consortium shed light on different aspects of cancers, such as mutational processes, tumor evolution, and diverse transcriptional consequences [82]. The combination of omics data from various databases such as Gene Expression Omnibus and TCGA is one of the promising approaches to finding potential biomarkers for understanding cancer development and progression [83].
Genetic variations such as single-nucleotide polymorphisms (SNPs) and deletion variants (indels) are associated with genetic susceptibility to cancer [84,85]. Pharmacogenetics studies explore genetic variations and focus on individual genetic variability and its correlation with drug efficacy, toxicity, and overall cancer treatment outcomes [86,87].
Epidermal growth factor receptor (EGFR) controls critical signaling pathways, including RAS/MAPK/ERK, PI3K/AKT, and STATS, and is responsible for cellular proliferation, differentiation, and survival. EGFR is an oncogenic driver in NSCLC, breast, and glioblastoma. EGFR has been linked to types of cancer, including metastatic colorectal cancer (CRC), pancreatic cancer, and head and neck squamous cell carcinoma (HNSCC). Its gene is located on the short arm of chromosome 7 at position 11.2 (7p11.2) [92]. EGFR is overexpressed and/or mutated in cancers such as glioblastoma, head and neck, pancreatic, breast, and metastatic colorectal cancer [93][94][95][96][97]. EGFR is overexpressed in 40-80% of NSCLC patients [98,99]. Abnormal EGFR expression has been associated with mutations in the EGFR, with approximately 10-15% in Caucasians with adenocarcinoma and 50% in lung tumors in East Asian populations [100]. Common mutations of EGFR are summarized in Figure 2. Several tyrosine kinase inhibitors (TKIs) targeting EGFR have been developed, including Erlotinib (Tarceva), Afatinib (Gilotrif), Gefitinib (Iressa), Osimertinib (Tagrisso), and Dacomitinib (Vizimpro), which improved progression-free survival (PFS), time-to-progression, and overall survival compared to standard chemotherapy in NSCLC patients. Gefitinib, Erlotinib, and Afatinib are common options for lung cancer patients with exon 21 Leu858Arg mutation or EGFR-exon 19 deletions. Unfortunately, patients with EGFR-mutant NSCLC develop disease progression within 10-14 months [101]. Different mechanisms underlie intrinsic or acquired resistance to EGFR-targeted inhibitors. T790M mutation in EGFR is responsible for approximately 50% of cases of acquired resistance [102]. Other important resistance mechanisms involve AKT mutations, loss of PTEN, activation of alternative signaling, and BCL2-like 11/BIM deletion polymorphism [103].
with exon 21 Leu858Arg mutation or EGFR-exon 19 deletions. Unfortunately, patients with EGFR-mutant NSCLC develop disease progression within 10-14 months [101]. Different mechanisms underlie intrinsic or acquired resistance to EGFR-targeted inhibitors. T790M mutation in EGFR is responsible for approximately 50% of cases of acquired resistance [102]. Other important resistance mechanisms involve AKT mutations, loss of PTEN, activation of alternative signaling, and BCL2-like 11/BIM deletion polymorphism [103]. The ALK gene consists of 30 exons and is located on chromosome 2p23. ALK is a tyrosine kinase receptor of the insulin receptor superfamily that plays an essential role in embryonic and neural development and human immune responses. The role of ALK in human tumorigenesis was discovered in anaplastic large cell lymphoma (ALCL) in 1994. The presence of EML4-ALK fusion was shown in breast, colorectal, and non-small-cell lung cancers. KIF5B, RET, ROS1, and ALK fusion has been detected in lung cancer. ALK could have different fusion partners, including NPM, in non-Hodgkin's lymphoma, VCL in Renal cell carcinoma, FN1 in gastrointestinal leiomyomas, and TRK-fused gene (TFG) in anaplastic large-cell lymphoma. The ALK gene consists of 30 exons and is located on chromosome 2p23. ALK is a tyrosine kinase receptor of the insulin receptor superfamily that plays an essential role in embryonic and neural development and human immune responses. The role of ALK in human tumorigenesis was discovered in anaplastic large cell lymphoma (ALCL) in 1994. The presence of EML4-ALK fusion was shown in breast, colorectal, and non-small-cell lung cancers. KIF5B, RET, ROS1, and ALK fusion has been detected in lung cancer. ALK could have different fusion partners, including NPM, in non-Hodgkin's lymphoma, VCL in Renal cell carcinoma, FN1 in gastrointestinal leiomyomas, and TRK-fused gene (TFG) in anaplastic large-cell lymphoma.
The V-RAF murine sarcoma viral oncogene homolog B (BRAF) encodes serine/threonine kinase. BRAF is a constituent of the MAPK pathway that regulates various cellular processes, including cell growth, proliferation, and survival. The effect of BRAF mutations in lung cancer patients was first reported in 2011 [104]. Mutations in BRAF are seen in 2-5 lung adenocarcinomas [105].
ROS1 (ROS proto-oncogene 1) is located on chromosome 6p22 and encodes 2347 amino acid residues. ROS1 is the subfamily of receptor tyrosine kinase, which regulates cell proliferation, survival, and migration [106]. FIG (GOPC)-ROS1 was the first ROS1 fusion gene reported in the human glioblastoma cell line U118MG in the early 1980s. Genetic rearrangements of ROS1 have been shown in different malignancies, including gastric adenocarcinoma, colorectal cancer, ovarian cancer, inflammatory myofibroblastic tumor, and angiosarcoma [106][107][108][109][110]. The first ROS1 fusion gene in NSCLC was identified in 2007. Previous studies have shown that ROS1 fusions account for 1-2% of all cases of NSCLC [88]. NSCLC patients with ROS1 fusion have several distinct clinical characteristics: they are typically female, younger (<50 years of age), and never or light smokers [111][112][113].

The PI3K Pathway in NSCLC
The phosphatidylinositol 3-kinase (PI3K)/Akt/mechanistic target of the rapamycin (mTOR) pathway has a pivotal role in regulating signal transduction and biological processes such as cell growth and survival, apoptosis, angiogenesis, tumor invasion, and metastasis. PI3Ks, AKT, and mTOR are the core components of the PI3K-AKT-mTOR signaling pathway, and their hyperactivation is observed in many cancers [114,115]. Therefore, alterations in this pathway lead to initiating and enhancing cancer progression. Consequently, this pathway is considered a target for novel anticancer therapies.
In NSCLC, the PI3K/Akt/mTOR pathway has been heavily implicated in carcinogenesis and disease advancement in NSCLC. Mutations in receptor tyrosine kinases (RTKs) and KRAS have been linked to the activation of the PI3K/Akt/mTOR axis in NSCLC [116,117]. Alterations in each component of this pathway may lead to lung cancer development and progression [118].  [120].
Recurrent alterations were observed in PIK3CA that were considered to contribute Osimertinib resistance [121]. A systematic review and meta-analysis study was conducted with a total of 13 studies involving 3908 NSCLC patients. The results indicated that PIK3CA mutations are associated with OS, PFS, and cancer-specific survival (CSS) [122]. PIK3CA mutation could be considered an independent prognostic factor for reduced PFS of EGFR-TKIs treatment and worse OS in NSCLC patients [122,123].
Several PI3K/Akt/mTOR signaling pathway inhibitors are currently developing as promising targets for new anticancer drugs in preclinical investigations and clinical trials. Pilaralisib is a highly selective inhibitor of the class I PI3Ks and successfully inhibits tumor growth in vivo. In phase I trials, Pilaralisib was assessed as a tolerable monotherapy in advanced solid tumors patients [109]. Alpelisib is a p110 alpha isoform-specific PI3K inhibitor that has been FDA-approved for HER2-positive advanced or metastatic breast cancer with PIK3CA-mutated. Alpelisib has been experimentally used in solid tumors with a favorable safety profile [109,124].
However, there is limited data on using PIK3CA inhibitors such as Copanlisib, Idelalisib, Umbralisib, Duvelisib, and Alpelisib in NSCLC patients. There is a need to elucidate the underlying molecular biology to detect the relevant biomarkers of toxicities and resistance to PI3K therapies. Several NGS-based panels related to cancer were approved by the FDA. Oncomine Dx Target Test detects variations in 23 genes that are biomarkers for selecting NSCLC patients for four targeted therapies Dabrafenib, Crizotinib, Gefitinib, and Pralsetinib [125]. The Praxis Extended RAS Panel detects several mutations in exons 2, 3, and 4 of KRAS and NRAS to implicate metastatic colorectal cancer treatment decisions [126]. MSK-IMPACT was created by the Memorial Sloan Kettering Cancer Center (MSK) to detect genetic aberrations in the 468-gene in solid tumors [127]. The FoundationOne CDx diagnostic test detects genetic mutations in 324 genes in several FDA-approved therapies for solid tumors [128]. Ongoing clinical trials are summarized in Table 4.
Of note, Crizotinib is an ALK, MET, and ROS1 kinases inhibitor. The phase I study of Crizotinib in 50 patients who were positive for ROS1 rearrangement proved the antitumor activity of this drug in advanced NSCLC [129]. Consequently, in 2016, Crizotinib was approved by the U.S. Food and Drug Administration (FDA) for the treatment of advanced NSCLC patients with ROS1-rearranged. In a phase II study of Crizotinib in East Asian patients, 127 East Asian patients were enrolled with ROS1-positive. The ORR by IRR was 71.7% (95% CI, 63.0% to 79.3) for NSCLC patients with ROS1-positive and had lower rates of brain metastases compared to ALK rearrangements (ROS1 19.4%, ALK 39.1%; p = 0.033) [130]. Another study identified 33 ROS1 rearrangements in 579 patients with stage IV NSCLC. The median PFS time for ROS1-positive was 11 months. ROS1 fusion partners were CD74, SLC34A2, and ZCCHC8 [131]. In Chinese patients with ROS1-positive advanced NSCLC treated with Crizotinib, the objective response rate (ORR) and the disease control rate (DCR) were 71.4% and 94.3%, respectively. In addition, the median PFS was 11.0 months (95% confidence interval (CI), 7.8-14.2), and median OS was 41.0 months (95% CI, 22.5-59.5) [132].

Current Status of Novel Biomarkers for Response to Immunotherapy
Immune checkpoint inhibitors (ICIs) are considered one of the most important treatments for various advanced malignancies and have become the standard of care in lung cancer [133,134]. Among ICI immunotherapy agents, the antiprogrammed death 1 ligand (PD-L1) (atezolizumab and durvalumab), the cytotoxic T-lymphocyte antigen 4 (anti-CTLA-4) (ipilimumab), and the programmed cell death receptor-1 (PD-1) (nivolumab and pembrolizumab) are currently used in therapy in patients with lung cancer. Treatment outcomes of lung cancer patients have improved considerably with these drugs; however, these drugs can cause immune-related adverse events (irAEs) that threaten the life of patients. Moreover, only a relatively small number of lung cancer patients benefit from immunotherapy, and the rapid tumor progression after the treatment with ICIs defined is as a hyperprogressive disease (HPD). Hence, an urgent need for biomarkers that predict responses and prognosis after treatment with ICIs is warranted.
Previous large clinical trials reported that the higher expression of PD-L1 can predict the response to PD1/PD-L1 inhibitors in different types of cancer, such as non-small-cell lung cancer (NSCLC), urothelial cancer, and melanoma [132,[135][136][137]. Immunohistochem-ical (IHC) detection of PD-L1 is a current predictive biomarker, but it is an imperfect biomarker for responses to ICI therapy.
In the KEYNOTE-001 trial, patients with advanced NSCLC enrolled in a phase I study and received Pembrolizumab. The PD-L1 expression in the tumor samples using the IHC analysis and results declared that at least 50% of the PD-L1 expression in the tumor cells correlated with improved efficacy of Pembrolizumab in patients [138]. In the open-label, phase II randomized controlled POPLAR trial, improvement of NSCLC patients treated with Atezolizumab was correlated with PD-L1 expression [139]. Data from the KEYNOTE-024 study reported that advanced NSCLC patients with PD-L1 expression on at least 50% of tumor cells treated with Pembrolizumab were associated with significantly longer PFS and OS [136]. However, there are some challenging obstacles to consistency in PD-L1 testing, including sampling methods, different techniques used to assess PDL1 expression, dynamic changes in PD-L1 expression, and spatial and temporal heterogeneity [122]. In addition, the single biomarker may not be indicative of patient selection. Therefore, several predictive and prognostic biomarkers have been explored during the past decade to guide patient choice.

Tumor Mutation Burden (TMB) and Circulating Tumor DNA (ctDNA)
The tumor mutation burden (TMB) refers to the total number of somatic mutations in tumor cells. The first association between a high TMB and treatment response was reported in patients with melanoma treated with cytotoxic T-lymphocyte-associated protein 4 (CTLA-4) [140]. Recently, various clinical trials showed that TMB is an important biomarker for predicting the efficacy of immunotherapy across diverse tumor types, including lung, urothelial, and breast cancers [141,142]. New studies have shown that blood-based TMB (bTMB) from circulating tumor DNA (ctDNA) has predictive power for the clinical benefit of immunotherapy in non-NSCLC patients receiving immunotherapy [142,143]. The role of bTMB profiling and sequencing of small amounts of cell-free DNA (cfDNA) in NSCLC patients treated with ICIs has been explored in several studies. Gandara et al. tested a novel assay to measure bTMB using samples collected prospectively from the POPLAR (NCT01903993) and the OAK (NCT02008227) clinical trials. Their results indicated that high bTMB in patients receiving Atezolizumab monotherapy in NSCLC was associated with longer PFS [143]. Wang et al. explored a 150-cancer genes panel (CGP) named NCC-GP150 in blood and tissue samples from non-small-cell lung cancer treated with anti-PD-1 and anti-PD-L1 therapy. Data from this analysis demonstrated that bTMB was a potential biomarker for patients with advanced NSCLC receiving ICIs [144]. Iijima et al. studied early response in 14 NSCLC patients treated with Nivolumab. The high level of tumor burden in circulating tumor DNA (ctDNA) could predict a durable response to Nivolumab [145]. Raja et al. studied variant allele frequencies (VAF) of somatic mutations in 73 genes ctDNA of NSCLC and urothelial cancer patients. Patients with the reduction in ctDNA VAF had a more significant decrease in tumor volume, with longer PFS and OS. [146]. In the study of Guibert et al., targeted sequencing of plasma ctDNA was analyzed in 97 progressive NSCLC patients. This study revealed that if a ctDNA allele fraction (A.F.) decreases, the PFS of patients increases. B-F1RST is the first prospective trial that evaluated bTMB in advanced NSCLC treated with Atezolizumab as a first-line (1L) monotherapy. The final result of this trial confirmed that bTMB could be considered a predictive biomarker for PFS and OS in NSCLC patients [146,147].
Atezolizumab monotherapy resulted in longer overall survival in NSCLC patients with high PD-L1 expression [148]. In the phase II B-F1RST (NCT02848651) trial, Kim et al. evaluated bTMB in locally advanced or metastatic stage IIIB-IVB NSCLC patients treated with first-line Atezolizumab monotherapy. In patients with bTMB ≥ 16, OS were associated with longer than patients with bTMB < 16 [142].
In meta-analyses of Meng et al., predictive value of TMB in NSCLC patients treated with ICIs were analyzed in six randomized controlled trials (3662 patients) and 31 datasets (3437 patients). Longer PFS, OS, and higher ORR were observed in high-TMB patients.
Furthermore, immunotherapy was associated with improved PFS, OS, and ORR in patients with high-TMB patients [149].
In the systematic review and meta-analysis of Kim et al., they explored the data of 5712 patients from 26 studies. The result of this study showed that, in patients who were treated with ICIs, the high-TMB groups showed better OS and PFS compared to the low-TMB groups. Moreover, patients with high TMB benefit more from ICI treatment than chemotherapy alone. In NSCLC patients, higher TMB was significantly associated with better PFS [150].
In the study of Thompson et al., ctDNA NGS was identified 50 drivers and 12 resistance mutations in advanced NSCLC. Detection of targetable driver and resistance mutations in ctDNA led to the identification of the resistance mechanisms in patients and to finding additional therapeutic options [151].

Proteomics
Protein markers can monitor treatment and/or predict drug resistance in NSCLC patients. Xu et al. utilized a SILAC-based quantitative proteomic approach to elucidate the Paclitaxel (PTX) resistance. They found that PDCD4 mediated PTX sensitivity in cancer cell lines, and during adjuvant therapy of lung cancer patients with PTX, the overall survival of patients correlated with high levels of PDCD4 [152]. Programmed cell death 4 (PDCD4) is downregulated in different cancer types and is considered a tumor suppressor. It has been proved that Pdcd4 has various functions inhibiting cell growth, suppressing tumor promotion, metastasis, and inducing apoptosis. For instance, the loss of PDCD4 expression is correlated with tumor progression and prognosis and survival in lung, breast, ovarian, and colorectal cancers. In lung cancer cells, Vihreva et al. demonstrated that the transcription of PDCD4 is negatively regulated by mTOR signaling.
Sandfeld-Paulsen et al. studied exosomal membrane proteins by using the extracellular vesicle array in 276 NSCLC patients. They found increasing concentration levels of NY-ESO-1, EGFR, PLAP, EpCam, and Alix have been associated with inferior OS [153]. In the same year, they published another study isolating exosomes from the blood of 581 patients (431 with lung cancer and 150 controls). They found that markers CD151, CD171, and tetraspanin eight could significantly determine NSCLC patients versus healthy controls [154]. In the PROSE study, Gregorc et al. demonstrated that a serum protein test could predict survival in non-small-cell lung cancer treated with Erlotinib and single-agent chemotherapy [155]. In the study by Salmon et al., L.C. patients were treated with Erlotinib in combination with Bevacizumab, and matrix-assisted laser desorption ionization time-of-flight mass spectrometry (MALDI-TOF MS) could classify patients who benefit from treatment with Erlotinib [155,156]. VeriStrat is a plasma-or serum-based test utilizing MALDI MS methods to monitor EGFR TKIs treatment of various tumor types, including colorectal and head and neck cancer [157]. Multiple studies demonstrated that the blood-based proteomic VeriStrat test could help predict EGFR TKI treatment response. In lung cancer, Taguchi et al. analyzed serum proteins by MALDI MS in NSCLC patients treated with EGFR TKIs. Proteomic evaluation of this study classified patients for good or poor outcomes after treatment with Gefitinib or Erlotinib [157,158]. Plasma samples of 111 NSCLC patients were analyzed by MALDI-TOF MS and ranked by the VeriStrat test [159]. Patients in the VeriStrat "good" classification had longer PFS and OS. Carbone [161]. In the study of Grossi et al., 481 NSCLC patients were treated with first-line platinum-based chemotherapy. In the group receiving VeriStrat, the PFS and OS were longer [161,162].

AI Machine Learning-Driven Discovery of Biomarkers for NSCLC
Robust clinical evaluation requires innovations, including technology platforms and artificial intelligence (AI). AI mimics the problem-solving and decision-making capabilities of the human mind [163]. Machine learning (ML) and its subfield deep learning are the main subsets of AI that can help biologists analyze vast amounts of data, leverage patterns in the data, and transform big data into actionable biomarkers [164]. AI revolutionized various areas in medicine, such as ophthalmology, radiology, and oncology [165][166][167]. Several medical algorithms detect CT-based lesions and are approved by the FDA, such as Arterys Oncology DL, Arterys MICA, and QuantX™, that improve radiologist performance [168,169]. AI affects different aspects of lung cancer, such as medical imaging for screening, early detection, characterizing lung cancers, treatment selection, and monitoring treatment response [170,171]. However, further improvement is required in big data interpretation and management, model fairness and interpretability, and generalizability.
Artificial intelligence has many applications in lung cancer [172]. Zhang et al. found that a combination of five urinary biomarkers discriminated lung cancer patients from healthy ones and differentiated lung cancer from other cancers [69]. Analysis of microbial reads of TCGA of 33 cancer types with the stochastic gradient-boosting machine (GBM) learning models successfully discriminated against cancer-free individuals, patients with cancer, and patients with multiple types of cancer [173]. In the study of Tirzïte et al., lung cancer patients' exhaled breath samples were explored with logistic regression analysis (LRA). This model can discriminate lung cancer patients from noncancer patients [174]. It is trained on a deep convolutional neural network (inception v3) that could predict most STK11, EGFR, FAT1, SETBP1, KRAS, and TP53 from pathology images [175]. The PD-L1 expression level is a crucial biomarker for identifying responders and nonresponders to NSCLC patients treated with anti-PD-1/PD-L1 treatments. However, the accurate estimation of PD-L1 expression is challenging. Wang et al. proposed a deep learning model that predicted mutated EGFR from the wild EGFR patients and distinguished PD-L1-positive from PD-L1-negative patients through C.T. images [176]. Choi et al. developed an AIpowered tumor proportion score (TPS) analyzer to evaluate whole-slide images of 802 NSCLC. Using this model, they could detect PD-L1 expression in tumors and calculate TPS. Moreover, AI-assisted TPS reading predicted OS and PFS upon ICI treatment [177]. Cheng et al. developed deep learning (DL)-based AI model to analyze the expression of PD-L1 in 1288 lung cancer patients. They used three different AI models (M1, M2, and M3) assessed in both PDL1 (22C3) and PD-L1 (SP263) assays. Their models improved the evaluation of PD-L1 expression, and diagnostic results were consistent with the pathologist's [177,178]. In the study of Wu et al., TPS of PD-L1 expression was also assessed in whole slide images (WSIs) of the 22c3 assay by the DL model. Their results also confirmed that AI-assisted diagnosis tests improve diagnostic repeatability and are a promising tool for improving the efficiency of clinical pathologists [179].

Conclusions and Future Directions
Cancer is a heterogeneous disease caused mainly by genomic aberrations. These aberrations may alter proteome, transcriptome, and metabolome pathways, which in turn become valuable biomarkers. Genomics and proteomics heterogeneity is the primary barrier that may be responsible for developing drug resistance. Using new techniques, such as artificial intelligence or single cell RNA sequencing, to solve old problems could solve old problems such as discovering biomarkers for diagnostic, prognostic, and predictive biomarkers to help guide clinicians to choose their medicine more precisely. These data layers integrate advanced computational methods to overview tumorigenesis comprehensively. Rethinking with new methods the resistance to checkpoint inhibitors (e.g., PD-1, PD-L1 or CTLA-4) or targeted therapies (e.g., PI3K/mROT pathways inhibitors) could help in selecting only patients who would benefit from these therapies in the future.
Solid high-throughput clinical outcomes are pivotal for clinical biomarkers discovery. Therefore, international collaboration with a large sample size of a wide variety of cancers and sharing of treatment data is imperative to develop more accurate biomarkers predicting the therapy response of lung cancer patients. Early detection and a proper treatment regimen prediction can ultimately aid clinicians in improving those patients' survival rates. Funding: This research received no external funding.

Conflicts of Interest:
The authors whose names are listed certify that they have no affiliations in any organization or entity with any financial interest and nonfinancial interest in the subject matter or materials discussed in this manuscript.