Identification of 13 Novel Loci in a Genome-Wide Association Study on Taiwanese with Hepatocellular Carcinoma.

Liver cancer is caused by complex interactions among genetic factors, viral infection, alcohol abuse, and metabolic diseases. We conducted a genome-wide association study and polygenic risk score (PRS) model in Taiwan, employing a nonspecific etiology approach, to identify genetic risk factors for hepatocellular carcinoma (HCC). Our analysis of 2836 HCC cases and 134,549 controls revealed 13 novel associated loci such as the FAM66C gene, noncoding genes, liver-fibrosis-related genes, metabolism-related genes, and HCC-related pathway genes. We incorporated the results from the UK Biobank and Japanese database into our study for meta-analysis to validate our findings. We also identified specific subtypes of the major histocompatibility complex that influence both viral infection and HCC progression. Using this data, we developed a PRS to predict HCC risk in the general population, patients with HCC, and HCC-affected families. The PRS demonstrated higher risk scores in families with multiple HCCs and other cancer cases. This study presents a novel approach to HCC risk analysis, identifies seven new genes associated with HCC development, and introduces a reproducible PRS model for risk assessment.

Immune responses and safety of COVID-19 vaccination in atypical hemolytic uremic syndrome patients in Taiwan.

Atypical hemolytic uremic syndrome is a rare, life-threatening disorder which can be triggered by COVID 19 infection and COVID 19 vaccination then induce multiple organ failure. Our study is the first to evaluate immune responses to COVID-19 vaccination and safety in a cohort of patients in a local single-center study in Taiwan.. Results indicate that vaccines effectively shield aHUS patients from severe COVID-19 complications without significant safety concerns. A double booster dose for the third vaccine is essential for optimal efficacy. Anti-complement therapy did not influence vaccination effectiveness. Transplant aHUS patients had the lowest immune response titers, indicating a need for additional vaccine doses. Compared to healthcare workers, aHUS patients had poor T-cell responses. We noted a superior trend with mixed-type COVID-19 vaccinations in aHUS patients, while fixed-type mRNA demonstrated better results in healthcare workers. Our findings endorse COVID-19 vaccination as a potent strategy to safeguard aHUS patients from severe complications, emphasizing the importance of vigilant monitoring pre- and post-vaccination.

Whole genome and RNA sequencing analyses for 254 Taiwanese hepatocellular carcinomas.

BACKGROUND: Comprehensive and integrative analysis of hepatocellular carcinoma (HCC) is important. In this study, we explored Taiwanese HCCs using multi-omics analyses. METHODS: We analyzed 254 HCCs by whole genome sequencing and total RNA sequencing, and then used bioinformatic tools to analyze genomic and transcriptomic alterations in coding and non-coding sequences to explore the clinical importance of each sequence. RESULTS: The frequencies of the five most commonly mutated cancer-related genes were TERT, TP53, CTNNB1, RB1, and ARID1A. Genetic alteration frequencies influenced the etiology of HCC; some alterations were also correlated with clinicopathological conditions. Many cancer-related genes had copy number alterations (CNAs) and structure variants (SVs) that changed according to etiology and exhibited potential associations with survival. We also identified several alterations in histone-related genes, HCC-related long non-coding RNAs, and non-coding driver genes that may contribute to the onset and progression of HCC. Transcriptomic analysis revealed that 229 differentially expressed and 148 novel alternative splicing (AS) genes, as well as the presence of fusion genes, were associated with patient survival. Moreover, somatic mutations, CNAs, and SVs were associated with immune checkpoint gene expression and tumor microenvironment. Finally, we identified relationships among AS, immune checkpoint gene expression and tumor microenvironment. CONCLUSIONS: This study shows that genomic alterations are associated with survival, including DNA-based and RNA-based data. Moreover, genomic alterations and their associations with immune checkpoint genes and the tumor microenvironment may provide novel insights for the diagnosis and treatment of HCC.

Comprehensive Analysis and Drug Modulation of Human Endogenous Retrovirus in Hepatocellular Carcinomas.

BACKGROUND: Human endogenous retroviruses (HERVs) play an important role in the development of cancer and many diseases. Here, we comprehensively explored the impact of HERVs on hepatocellular carcinomas (HCCs). METHODS: We employed Telescope to identify HERVs and quantify their expression in the total RNA sequencing data obtained from 254 HCC samples, comprising 254 tumor tissues and 34 matched normal tissues. RESULTS: In total, 3357 locus-specific activations of HERVs were differentially expressed, and 180 were correlated with patient survival. Using these 180 HERVs for classification, we found four subgroups with survival correlation. Higher expression levels of the 180 HERVs were correlated with poorer survival, while age, AFP, some mutations, and copy and structural variants differed among subgroups. The differential expression of host genes in high expression of these 180 HERVs primarily involved the activation of pathways related to immunity and infection, lipid and atherosclerosis, MAPK and NF-kB signaling, and cytokine-cytokine receptor interactions. Conversely, there was a suppression of pathways associated with RNA processing, including nucleocytoplasmic transport, surveillance and ribosome biogenesis, and transcriptional misregulation in cancer pathways. Almost all genes involved in HERV activation restriction, KRAB zinc finger proteins, RNA nucleocytoplasmic transport, stemness, HLA and antigen processing and presentation, and immune checkpoints were overexpressed in cancerous tissues, and many over-expressed HERV-related nearby genes were correlated with high HERV activation and poor survival. Twenty-three immune and stromal cells showed higher expression in non-cancerous than cancerous tissues, and seven were correlated with HERV activation. Small-molecule modulation of alternative splicing (AS) altered the expression of survival-related HERVs and their activation-related genes, as well as nearby genes. CONCLUSION: Comprehensive and integrated approaches for evaluating HERV expression and their correlation with specific pathways have the potential to provide new companion diagnostics and therapeutic strategies for HCC.

Whole exome sequencing identifies MAP3K1, MSH2, and MLH1 as potential cancer-predisposing genes in familial early-onset colorectal cancer.

The incidence of early-onset colorectal cancer (CRC), which affects people under 50, is increasing for unknown reasons. Additionally, no underlying genetic cause is found in 20%-30% of patients suspected of having familial CRC syndrome. Whole exome sequencing (WES) has generated evidence for new genes associated with CRC susceptibility, but many patients remain undiagnosed. This study applied WES in five early-onset CRC patients from three unrelated families to identify novel genetic variants that could be linked to rapid disease development. Furthermore, the candidate variants were validated using Sanger sequencing. Two heterozygote variations, c.1077-2A>G and c.199G>A, were found in the MSH2 and the MLH1 genes, respectively. Sanger sequencing analysis confirmed that these (likely) pathogenic mutations segregated in all the affected families' members. In addition, we identified a rare heterozygote variant (c.175C>T) with suspected pathogenic potential in the MAP3K1 gene; formally the variant is of uncertain significance (VUS). Our findings support the hypothesis that CRC onset may be oligogenic and molecularly heterogeneous. Larger and more robust studies are needed to understand the genetic basis of early-onset CRC development, combined with novel functional analyses and omics approaches.

Dysregulation of Immune Cell Subpopulations in Atypical Hemolytic Uremic Syndrome.

Atypical hemolytic uremic syndrome (aHUS) is a rare, life-threatening thrombotic microangiopathy. Definitive biomarkers for disease diagnosis and activity remain elusive, making the exploration of molecular markers paramount. We conducted single-cell sequencing on peripheral blood mononuclear cells from 13 aHUS patients, 3 unaffected family members of aHUS patients, and 4 healthy controls. We identified 32 distinct subpopulations encompassing 5 B-cell types, 16 T- and natural killer (NK) cell types, 7 monocyte types, and 4 other cell types. Notably, we observed a significant increase in intermediate monocytes in unstable aHUS patients. Subclustering analysis revealed seven elevated expression genes, including NEAT1, MT-ATP6, MT-CYB, VIM, ACTG1, RPL13, and KLRB1, in unstable aHUS patients, and four heightened expression genes, including RPS27, RPS4X, RPL23, and GZMH genes, in stable aHUS patients. Additionally, an increase in the expression of mitochondria-related genes suggested a potential influence of cell metabolism on the clinical progression of the disease. Pseudotime trajectory analysis revealed a unique immune cell differentiation pattern, while cell-cell interaction profiling highlighted distinctive signaling pathways among patients, family members, and controls. This single-cell sequencing study is the first to confirm immune cell dysregulation in aHUS pathogenesis, offering valuable insights into molecular mechanisms and potential new diagnostic and disease activity markers.

Highly Mimetic Ex Vivo Lung-Cancer Spheroid-Based Physiological Model for Clinical Precision Therapeutics.

Lung cancer remains a major health problem despite the considerable research into prevention and treatment methods. Through a deeper understanding of tumors, patient-specific ex vivo spheroid models with high specificity can be used to accurately investigate the cause, metastasis, and treatment strategies for lung cancer. Biofabricate lung tumors are presented, consisting of patient-derived tumor spheroids, endothelial cells, and lung decellularized extracellular matrix, which maintain a radial oxygen gradient, as well as biophysicochemical behaviors of the native tumors for precision medicine. It is also demonstrated that the developed lung-cancer spheroid model reproduces patient responses to chemotherapeutics and targeted therapy in a co-clinical trial, with 85% accuracy, 86.7% sensitivity, and 80% specificity. RNA sequencing analysis validates that the gene expression in the spheroids replicates that in the patient's primary tumor. This model can be used as an ex vivo predictive model for personalized cancer therapy and to improve the quality of clinical care.

Integrated genomic analyses of hepatocellular carcinoma.

BACKGROUND: Genomic alterations play important roles in the development of cancer. We explored the impact of protein-coding genes and transcriptomic changes on clinical and molecular alterations in Taiwanese hepatocellular carcinoma (HCC) patients. METHODS: We analyzed 147 whole-exome sequencing and 100 RNA sequencing datasets of HCC and compared them with The Cancer Genome Atlas (TCGA)-Liver Hepatocellular Carcinoma cohort and develop a panel of 81 apoptosis-related genes for molecular classification. RESULTS: TERT (50%), TP53 (25%), CTNNB1 (14%), ARID1A (12%), and KMT2C (11%) were the most common genetic alterations of cancer-related genes. ALDH2 and KMT2C mutated at much higher frequencies in our cohort than in TCGA, whereas CTNNB1 was found only in 14% of our Taiwanese patients. A high germline mutation rate of ALDH2 in the APOBEC mutational signature and herb drug-related aristolochic acid-associated signature was also observed. Groups A and B of HCC were identified when we used apoptosis-related genes for molecular classification. The latter group, which had poorer survival outcomes, had significantly more aDC, CD4+ Tem, macrophages M2, NKT, plasma cells, and Th1 cells, and less CD4+ memory T cells, CD8+ Tcm, cDC, iDC, and Th2 cells, as well as more inter-chromosome fusion genes. Metatranscriptomic analysis revealed 54 cases of HBV infection. Moreover, we found that the main target gene of HBV integration is ALB. CONCLUSIONS: Unique genomic alterations were observed in our Taiwanese HCC patients. Molecular classification using apoptosis-related genes could lead to new therapeutic approaches for HCC.

Cancer carrier screening in the general population using whole-genome sequencing.

BACKGROUND: Cancer is a major cause of death, and its early identification and intervention have potential for clinical actionability and benefits for human health. The studies using whole-genome sequencing (WGS) and large samples analysis of cancer-related genes have been rarely done. METHODS: We performed WGS to explore germline mutations in coding and non-coding areas of cancer-related genes and non-coding driver genes and regulatory areas. Structural variants (SVs) was also analyzed. We used several tools and a subgrouping method to analyze the variants in 1491 healthy participants. Moreover, 275 cancer-related genes sequencing was carried out in 125 cancer patients. RESULTS: The incidence of familial cancer in the Taiwanese general population is 8.79% (131/1491). Cancer carrier rate of cancer-related genes is about 7.04% (105/1491) for pathogenic/likely pathogenic variants (P/LP) on ClinVar database only, and 28.24% (421/1491) for P/LP and loss of function variants. The carrier frequencies of cancer-related genes P/LP on ClinVar database were as follows: 8.40% (11/131), 7.11% (28/394), and 6.83% (66/966) in FC, 1MC, and nMC, respectively. The SVs and non-coding driver gene variants are uncommon. There are 1.54% (23/1491) of actionable cancer genes in American College of Medical Genetics and Genomics (ACMG), and the germline mutation rate of 275 cancer-related genes is 7.2% (9/125) in cancer patients including 4.0% (5/125) of actionable cancer genes in ACMG. After analyzing the frequencies of P/LP variants on GJB2 and SLC25A13 genes, we suggest that these two genes may not be cancer-related genes and need be re-evaluated. CONCLUSIONS: WGS analysis can completely detect germline mutations in cancer carriers. This study use subgrouping approach for samples provides a strategy to study whether a gene or variant is a cancer-related gene or variant in the future studies.

Genetic contributions to female gout and hyperuricaemia using genome-wide association study and polygenic risk score analyses.

OBJECTIVES: To identify genetic variants and polygenic risk score (PRS) relating to female gout and asymptomatic hyperuricaemia (AH) in a genome-wide association study (GWAS). METHODS: Gout, AH and normouricemia controls were included from Taiwan biobank and China Medical University Hospital. All participants were divided into discovery and replication cohorts for GWAS. PRS was estimated according to whether the variant exhibited a protective effect on the phenotypes or not. Each cohort was separated into two groups by the age of 50 years old. RESULTS: A total of 59 472 females were enrolled, and gout and AH occupied 1.60% and 19.59%, respectively. Six variants located in genes SLC2A9, C5orf22, CNTNAP2 and GLRX5 were significantly predictors of female gout in those aged ≥50. For those aged <50 years old, only the variant rs147750368 (SPANXN1) on chromosome X was found. Most variants located in genes SLC2A9, ZNF518B, PKD2 and ABCG2 were found to be significantly related to AH in both age groups. The PRS could explain ∼0.59% to 0.89% of variance of gout in variants with protective effects, which showed 6.2 times of mean PRS in the risk variants, but only 1.2 times in the AH phenotype. Moreover, the PRS also revealed a dose-response trend between AH rates and quartile scores. CONCLUSION: The variants in gene SLC2A9 are the major genetic factors for females associated with gout in those aged ≥50. PRS can provide a more robust prediction of the gout/AH under a homogeneous selection of variants that show effects on the traits.

Single-cell RNA sequencing of immune cells in patients with acute gout.

Cell subpopulations in the blood and joint fluid of patients with gout are poorly understood. Single-cell RNA sequencing and bioinformatic tools were used to identify cell subsets and their gene signatures in blood and synovial fluid (SF) cells, determine their relationships, characterize the diversity, and evaluate interactions among specific cell types. We identified 34 subpopulations (5 types of B cells, 16 types of T and natural killer cells, 9 types of monocytes, and 4 other cell types) in the blood of five healthy subjects and seven patients with acute gouty, and the SF of three patients with acute gout. We found that naïve CD4 T cells and classical monocytes cell populations were enriched in patients with gout, whereas plasmacytoid dendritic cells and intermediate monocytes were more abundant in healthy subjects. SF was enriched in Th1/Th17 cells, effector memory CD8 T cells, mucosal-associated invariant T cells, and macrophages. Subclusters of these cell subpopulations showed different compositions between healthy subjects and those with acute gout, according to blood and SF samples. At the cellular level, the inflammation score of a subpopulation or subcluster was highest in SF, following by the blood of acute gout patients and healthy person, whereas energy score showed the opposite trend. We also detected specific cell-cell interactions for interleukin-1, tumor necrosis factor-α, and transforming growth factor-ß1 expression in the cells of patients with acute gout. Our study reveals cellular and molecular insights on inflammatory responses to hyperuricemia or uric crystal and may provide therapeutic guidance to improve treatments for gout.

Polygenic risk score trend and new variants on chromosome 1 are associated with male gout in genome-wide association study.

BACKGROUND: Gout is a highly hereditary disease, but not all those carrying well-known risk variants have developing gout attack even in hyperuricemia status. We performed a genome-wide association study (GWAS) and polygenic risk score (PRS) analysis to illustrate the new genetic architectures of gout and asymptomatic hyperuricemia (AH). METHODS: GWAS was performed to identify variants associated with gout/AH compared with normouricemia. The participants were males, enrolled from the Taiwan Biobank and China Medical University, and divided into discovery (n=39,594) and replication (n=891) cohorts for GWAS. For PRS analysis, the discovery cohort was grouped as base (n=21,814) and target (n=17,780) cohorts, and the score was estimated by grouping the polymorphisms into protective or not for the phenotypes in the base cohort. RESULTS: The genes ABCG2 and SLC2A9 were found as the major genetic factors governing gouty and AH, and even in those carrying the rs2231142 (ABCG2) wild-genotype. Surprisingly, variants on chromosome 1, such as rs7546668 (DNAJC16), rs10927807 (AGMAT), rs9286836 (NUDT17), rs4971100 (TRIM46), rs4072037 (MUC1), and rs2974935 (MTX1), showed significant associations with gout in both discovery and replication cohorts (all p-values < 1e-8). Concerning the PRS, the rates of gout and AH increased with increased quartile PRS in those SNPs having risk effects on the phenotypes; on the contrary, gout/AH rates decreased with increased quartile PRS in those protective SNPs. CONCLUSIONS: We found new variants on chromosome 1 significantly relating to gout, and PRS predicts the risk of developing gout/AH more robustly based on the SNPs' effect types on the trait.

Molecular Classification of Hepatocellular Carcinoma Using Wnt-Hippo Signaling Pathway-Related Genes.

In Taiwan, a combination of hepatitis B and C infection, economic boom-related food and alcohol overconsumption, and Chinese medicine prescriptions has led to a high rate of hepatocellular carcinoma (HCC). However, the causative factors and underlying tumor biology for this unique HCC environment have not been identified. Wnt and Hippo signaling pathways play an important regulatory role in HCC development, and their functions are generally considered as positive and negative regulators of cell proliferation, respectively. In this study, we characterized the molecular features of HCC using a newly developed classification system based on the expression of the Wnt-Hippo signaling pathway-related genes. RNA sequencing (RNA-Seq) was performed on liver tumor tissues from 100 patients with liver cancer. RNA-Seq data for 272 previously characterized Wnt-Hippo signaling pathway-related genes were used for hierarchical clustering. We analyzed the data in terms of prognostic value, transcriptome features, immune infiltration, and clinical characteristics, and compared the resulting subclasses with previously published classifications. Four subclasses of HCC (HCCW1-4) were identified. Subclass HCCW1 displayed the highest PCDHB4 expression. Subclass HCCW2 displayed lower Edmondson-Steiner grades (I and II) and CTNNB1 mutation frequencies. Subclass HCCW3 was associated with a good prognosis, the highest PCDHGB7 expression, high CD8+ naïve T cells abundance, and relatively low TP53 mutation rates. Subclass HCCW4 was associated with a poor prognosis, the highest PCDHB2 and PCDHB6 expression, a relatively high abundance of Th1 cells, NKT and class-switched memory B cells, relatively low enrichment of cDC, iDC, and CD4+ memory T cells, and high Edmondson-Steiner grades (III and IV). We also identified Wnt-Hippo signaling pathway-related genes that may influence immune cell infiltration. We developed a panel of 272 Wnt-Hippo signaling pathway-related genes to classify HCC into four groups based on Taiwanese HCC and The Cancer Genome Atlas Liver Hepatocellular Carcinoma datasets. This novel molecular classification system may aid the treatment of HCC.

Comprehensive characterization of pharmacogenes in a Taiwanese Han population.

Pharmacogenetic (PGx) testing has not been well adopted in current clinical practice. The phenotypic distribution of clinically relevant pharmacogenes remains to be fully characterized in large population cohorts. In addition, no study has explored actionable PGx alleles in the East Asian population at a large scale. This study comprehensively analyzed 14 actionable pharmacogene diplotypes and phenotypes in 172,854 Taiwanese Han individuals by using their genotype data. Furthermore, we analyzed data from electronic medical records to investigate the effect of the actionable phenotypes on the individuals. The PGx phenotype frequencies were comparable between our cohort and the East Asian population. Overall, 99.9% of the individuals harbored at least one actionable PGx phenotype, and 29% of them have been prescribed a drug to which they may exhibit an atypical response. Our findings can facilitate the clinical application of PGx testing and the optimization of treatment and dosage individually.

Exploring RNA modifications, editing, and splicing changes in hyperuricemia and gout.

Hyperuricemia and gout are two of the most common metabolic disorders worldwide; their incidence is increasing with changes in lifestyle, and they are correlated with many diseases, including renal and cardiovascular diseases. The majority of studies on hyperuricemia and gout have focused on the discovery of the associated genes and their functions and on the roles of monocytes and neutrophils in the development of gout. Virtually no studies investigating the epigenomics of gout disease or exploring the clinical significance of such research have been conducted. In this study, we observed that the expression of enzymes involved in RNA modifications or RNA editing was affected in uric acid (UA)- or monosodium urate (MSU)-treated cell lines. RNA alternative splicing and splicing factors were also affected by UA or MSU treatment. We used transcriptome sequencing to analyze genome-wide RNA splicing and RNA editing and found significant changes in RNA splicing and RNA editing in MSU- or UA-treated THP-1 and HEK293 cells. We further found significant changes of RNA modifications, editing, and splicing in patients with gout. The data indicate that RNA modifications, editing, and splicing play roles in gout. The findings of this study may help to understand the mechanism of RNA splicing and modifications in gout, facilitating the development of new diagnostic and therapeutic strategies.

Genome-wide association study of hyperthyroidism based on electronic medical record from Taiwan.

Excess thyroid hormones have complex metabolic effects, particularly hyperthyroidism, and are associated with various cardiovascular risk factors. Previous candidate gene studies have indicated that genetic variants may contribute to this variable response. Electronic medical record (EMR) biobanks containing clinical and genomic data on large numbers of individuals have great potential to inform the disease comorbidity development. In this study, we combined electronic medical record (EMR) -derived phenotypes and genotype information to conduct a genome-wide analysis of hyperthyroidism in a 35,009-patient cohort in Taiwan. Diagnostic codes were used to identify 2,767 patients with hyperthyroidism. Our genome-wide association study (GWAS) identified 44 novel genomic risk markers in 10 loci on chromosomes 2, 6, and 14 (P < 5 × 10-14), including CTLA4, HCP5, HLA-B, POU5F1, CCHCR1, HLA-DRA, HLA-DRB9, TSHR, RPL17P3, and CEP128. We further conducted a comorbidity analysis of our results, and the data revealed a strong correlation between hyperthyroidism patients with thyroid storm and stroke. In this study, we demonstrated application of the PheWAS using large EMR biobanks to inform the comorbidity development in hyperthyroidism patients. Our data suggest significant common genetic risk factors in patients with hyperthyroidism. Additionally, our results show that sex, body mass index (BMI), and thyroid storm are associated with an increased risk of stroke in subjects with hyperthyroidism.

Divergent Characteristics of T-Cell Receptor Repertoire Between Essential Hypertension and Aldosterone-Producing Adenoma.

Aldosterone-producing adenoma (APA) is a benign adrenal tumor that results in persistent hyperaldosteronism. As one major subtype of primary aldosteronism, APA leads to secondary hypertension that is associated with immune dysregulation. However, how the adaptive immune system, particularly the T-cell population, is altered in APA patients remains largely unknown. Here, we performed TCR sequencing to characterize the TCR repertoire between two age-matched groups of patients: one with APA and the other one with essential hypertension (EH). Strikingly, we found a significant reduction of TCR repertoire diversity in the APA group. Analyses on TCR clustering and antigen annotation further showed that the APA group possessed lower diversity in TCR clonotypes with non-common antigen-specific features, compared with the EH group. In addition, our results indicated that the strength of correlation between generation probabilities and frequencies of TCR clonotypes was significantly higher in the APA group than that in the EH group. Finally, we observed that clinical features, including plasma aldosterone level, aldosterone-renin ratio, and blood sodium level, were positively associated with the strength of correlation between generation and abundance of TCR clonotypes in the APA group. Our findings unveiled the correlation between T-cell immune repertoire and APA, suggesting a critical role of such adrenal adenoma in the T-cell immunity of patients with hypertension.

Mutation Analysis of Second Primary Tumors in Oral Cancer in Taiwanese Patients through Next-Generation Sequencing.

Head and neck cancer has poor overall survival. Patients with head and neck cancer more frequently develop second primary tumors than do patients with other cancers, leading to a poor prognosis. In this study, we used next-generation sequencing to analyze and compare mutations between first tumors and second tumors in oral cancer. We retrieved tumor tissues collected from 13 patients who were diagnosed twice as having cancer. We used driver gene and trunk mutations to distinguish between recurrent cancer and primary cancer in oral cancer. We observed unique driver gene mutations in three patients with an initial clinical diagnosis of recurrent cancer; hence, we believe that the corresponding patients had primary cancer. Four patients with an initial clinical diagnosis of primary cancer were found to actually have recurrent cancer according to our results. Genetic testing can be used to enhance the accuracy of clinical diagnosis.

A nine-gene signature identification and prognostic risk prediction for patients with lung adenocarcinoma using novel machine learning approach.

BACKGROUND: Lung adenocarcinoma (LUAD) is one the most prevalent cancer with high mortality and its risk stratification is limited due lack of reliable molecular biomarkers. Although several studies have been conducted to identify gene signature involved in LUAD progression, most currently used methods to select gene features did not fully consider the problem of the existence of strong pairwise gene correlations as it resulted inconsistency in gene election. Therefore, it is crucial to develop new strategy to identify reliable gene signatures that improve risk prediction. METHODS AND RESULTS: In this study, novel feature selection strategy (1) univariate Cox regression model to select survival associated genes (2) integrating rigid Cox regression with Adaptive Lasso model to identify informative survival associated genes (3) stepwise Cox regression model to identify optimal gene signature and (4) prognostic risk predictive model for LUAD (PRPML) was constructed. The PRPML was developed-based on four machine learning (ML) methods including logistic regression (LR), K-nearest neighbors (KNN), support vector machine with the radial kernel (SVMR), and average neural network (Avnet). The PRPML model successfully stratified high-risk and low-risk groups of patients with LUAD in three datasets. The PRPML achieved an area under the curve (AUC) of 0.812 and 0.863 in the validation datasets. Finally, a nine-potential gene signature was found and showed great potential for risk prediction. CONCLUSIONS: Our study demonstrates that the developed strategy identified a nine potential gene signature for accurate risk prediction performance and this signature could provide valuable clue into the understanding of the molecular mechanism of LUAD disease.