Finally, our study selects the optimal gene-expression-signature for the prediction of cancer risk via quantitative real-time PCR via LASSO regression. CD19 is an essential receptor for B cell antigen receptor (BCR) signal transduction. Co-ligation of CD19 could enhance mitogen-activated protein kinase activity and cell proliferation, and could also negatively regulate BCR signaling. Anti-CD19 chimeric antigen receptor T cells are currently used in transformational therapy for aggressive B-cell lymphomas [
47,
48]. CDKN1A (P21) is a member of cyclin-dependent kinase inhibitors and it has been regarded as a tumor suppressor by regulating the cell cycle and maintaining genomic stability. The downregulation of CDKN1A is linked to poor prognosis in multiple cancers [
49]. However, its overexpression is also found in a variety of human cancers [
49]. Moreover, the upregulation of CDKN1A and its frequently cytoplasmic relocation correlate positively with poor prognosis in gastric cancer. The role of CDKN1A on tumorigenesis depends on the cellular context, its subcellular localization and posttranslational modifications. The application of CDKN1A as a prognostic marker and a therapeutic target in cancer still require further investigation [
49]. S100A9 (calgranulin B, Calprotectin) is a Ca(2 +) binding protein involved in inflammatory processes. S100A9 was elevated in inflammation and various human cancers[
50]. Recent studies have shown the presence of S100A9 and inflammatory factors in the tumor microenvironment. The function of S100A9 depends on its concentration and location. S100A9 at high extracellular concentrations could induce the apoptosis pathway in cancer cells, while at lower levels S100A9 seem to promote proliferation of tumor cells [
50]. However, at high intracellular concentrations, S100A9 induces a reduction in cancer cell invasion capacity by regulating the epithelial–mesenchymal transition–the mesenchymal–epithelial transition (EMT–MET) signaling cascades [
50]. The molecular mechanism of pro- and anti-tumor properties of S100A9 is still unknown. TAP1 is associated with antigen processing of major histocompatibility complex class I peptides for recognition by tumor-specific cytotoxic T lymphocytes. TAP1 overexpression might be an indicator of aggressive breast cancer and was also significantly associated with poor prognosis in colorectal cancer [
51‐
53]. Moreover, decreased TAP1 protein expression was significantly associated with low infiltration of lymphocytes and macrophages [
51,
52]. Bioinformatic study with large datasets demonstrated a correlation between the TAP1 gene and tumor progression and a significant negative correlation for TAP1 gene expression and the survival rate in different cancer types [
53]. TNFRSF1B belongs to TNF receptor (TNFR) superfamilies and plays an important role in protective immunity, inflammatory and tumor immunology. TNFRSF1B can induce downstream signaling pathways such as NF-κB and PI3K/Akt activation when interacted with its ligand TNFα [
54]. TNFα-mediated co-stimulation supports TCR/CD28-mediated T cell activation and survival [
54]. Furthermore, ligation of TNFRSF1B inhibits regulatory T cell differentiation by suppressing Smad3-dependent Foxp3 transcription [
54]. CCR7 when ligated with its ligands, could induce the homing of T cells to a lymph node. Therefore, the increased expression of CCR7 has an anti-cancer effect via cytotoxic TIL in tumors [
55]. However, CCR7 could also enhance proliferation and stemness of cancer cells. The mechanisms of the tumor-promoting effect of CCR7 include the induction of tumor angiogenesis by activating NF-κB and increasing VEGF expression, epithelial–mesenchymal transition of cancer cells and migration of cancer cells to metastasis sites [
55]. Higher expression of CCR7 is also associated with worse prognosis in diffuse large B-cell lymphoma [
55]. CD4 is a coreceptor with the T-cell receptor on the T lymphocyte and CD4( +) T cell help signals are relayed to CD8( +) T cells by specific dendritic cells to optimize cytotoxic T lymphocyte (CTL) response [
56]. Deficient CD4( +) T cell help can reduce the response of CTLs and maximizing CD4( +) T cell help can improve outcomes in cancer immunotherapy [
56]. KMT2D belongs to the lysine methyltransferase (KMT2) family of histone modifying proteins, which play essential roles in regulating developmental pathways. The KMT2A-D proteins are important for RNA Polymerase II-dependent transcription and KMT2 mutations have been linked to multiple cancers [
57]. Recent studies have also provided evidence for KMT2 protein participation in epigenetic gene regulation and in carcinogenesis [
57]. LY6E belongs to the LY6 gene family, which represent novel biomarkers for poor cancer prognosis and play essential roles in cancer progression and immune escape [
58]. LY6E expressions are increased in bladder cancer, gastric cancer, etc. [
58]. The LY6E gene has also been associated with more aggressive stem like cells in hepatocellular carcinoma, pancreatic carcinoma, etc. [
58]. Recent data suggest that increased expression of LY6E is associated with poor overall survival of renal papillary cell carcinoma and pancreatic ductal adenocarcinoma [
58]. These genes in our final gene signature are mostly protein coding genes involved in cancer immunology. It is possible that platelet mRNA could be used in cancer risk prediction in other types of cancers.
Indeed there are previous longitudinal studies using pre-diagnostic serums to screen for novel biomarkers of early cancer detection. However, these studies utilized pre-diagnostic serum to detect tumor-specific antigens or auto-antibodies for cancer risk prediction with limited sensitivity [
7‐
10]. Novel cancer markers such as circulating tumor cells (CTCs) and circulating tumor DNA (ctDNA) offer new genomic approaches to screen for cancer through liquid biopsies. However, recent studies indicate CTCs assay cannot differentiate between patients with early-stage malignancy and people with no cancer and it has limited specificity as a screening tool [
3,
4]. On the other hand, ctDNA has promised to be a sensitive and specific test for cancer screening [
11,
12]. Still, ctDNA testing has several limitations for a screening platform compared with platelet RNA testing. First, the quantity of ctDNA is very limited even in cancer patients, not to mention in patients with early-stage cancer, while blood platelets are quite abundant. So the volume of blood needed in platelet testing is about 0.1 ml while ctDNA testing requires at least 10 ml. Second, the isolation and conversion process may cause damages to ctDNA, while platelet isolation procedure is simple and sample is stable and easy for storage. Third, ctDNA extraction requires an expensive kit while platelet isolation needs no expensive consumables. The subsequent sequencing analysis of ctDNA is also more expensive than platelet sequencing in our study. Furthermore, our study used LASSO regression to select the optimum gene-expression-signature for the prediction of cancer risk via quantitative real-time PCR. Hence our strategy with the prediction models including 4 or 5 biomarkers as variables is much more cost-effective than ctDNA testing for hundreds of hotspots. Fourth, ctDNA analysis could only detect frequently mutated genes in common cancers. The evolutionary and heterogeneity nature of cancer demands a large amount of possible mutations to be screened to achieve a consistent biomarker. Platelet biomarkers, on the other hand, are genes correlated with immune response and regulation according to our findings. Hence, platelet RNA testing may not be affected by cancer type or heterogeneity. Fifth, our platelet RNA prediction model could discriminate early-stage cancer from both healthy control and macroscopic tumor group, while biomarkers or screening models from previous studies often cannot distinguish samples from different stages of cancer. Thus platelet RNA testing may easily determine the best window for possible intervention. Last but not the least, platelet RNA testing described in our proof-of-concept study takes hours via qPCR while ctDNA testing takes days or weeks via next-generation sequencing and require skilled biology and bioinformatics technicians. Hence platelet testing is much less time-consuming and requires less training of technicians. This demonstrates the potential of platelets as a non-invasive screening platform for the detection of occult cancer.