Identifying microRNAs associated with tumor immunotherapy response using an interpretable machine learning model.

Predicting clinical responses to tumor immunotherapy is essential to reduce side effects and the potential for sustained clinical responses. Nevertheless, preselecting patients who are likely to respond to such treatments remains highly challenging. Here, we explored the potential of microRNAs (miRNAs) as predictors of immune checkpoint blockade responses using a machine learning approach. First, we constructed random forest models to predict the response to tumor ICB therapy using miRNA expression profiles across 19 cancer types. The contribution of individual miRNAs to each prediction process was determined by employing SHapley Additive exPlanations (SHAP) for model interpretation. Remarkably, the predictive performance achieved by using a small number of miRNAs with high feature importance was similar to that achieved by using the entire miRNA set. Additionally, the genes targeted by these miRNAs were closely associated with tumor- and immune-related pathways. In conclusion, this study demonstrates the potential of miRNA expression data for assessing tumor immunotherapy responses. Furthermore, we confirmed the potential of informative miRNAs as biomarkers for the prediction of immunotherapy response, which will advance our understanding of tumor immunotherapy mechanisms.

Distinct cellular composition between normal surgical margins and tumor tissues in oral squamous cell carcinoma.

BACKGROUND: Adequate resection of normal surgical margins is important. However, the clear distinction between the normal surgical margins and tumor tissues is still difficult. OBJECTIVE: Here, this study analyzed the variety of cell types in tumors and the normal surgical margins using a computational approach. METHODS: The composition of cell types was compared between the two tissues by statistical and machine learning approaches. RESULTS: The results showed the distinct cellular composition between tumor-adjacent and tumor tissues. In particular, endothelial cells were highly represented and macrophages were underrepresented at the normal surgical margin. Moreover, the normal surgical margin and tumor tissues could be discriminated using a machine learning algorithm. CONCLUSION: The results will help to understand cellular differences between normal surgical margins and tumor tissues and to provide potentials for tumor detection and treatment.

Genomic Effect of DNA Methylation on Gene Expression in Colorectal Cancer.

The aberrant expression of cancer-related genes can lead to colorectal cancer (CRC) carcinogenesis, and DNA methylation is one of the causes of abnormal expression. Although many studies have been conducted to reveal how DNA methylation affects transcription regulation, the ways in which it modulates gene expression and the regions that significantly affect DNA methylation-mediated gene regulation remain unclear. In this study, we investigated how DNA methylation in specific genomic areas can influence gene expression. Several regression models were constructed for gene expression prediction based on DNA methylation. Among these models, ElasticNet, which had the best performance, was chosen for further analysis. DNA methylation near transcription start sites (TSS), especially from 2 kb upstream to 7 kb downstream of TSS, had an essential regulatory role in gene expression. Moreover, methylation-affected and survival-associated genes were compiled and found to be mainly enriched in immune-related pathways. This study investigated genomic regions in which methylation changes can affect gene expression. In addition, this study proposed that aberrantly expressed genes due to DNA methylation can lead to CRC pathogenesis by the immune system.

Assessment of MicroRNAs Associated with Tumor Purity by Random Forest Regression.

Tumor purity refers to the proportion of tumor cells in tumor tissue samples. This value plays an important role in understanding the mechanisms of the tumor microenvironment. Although various attempts have been made to predict tumor purity, attempts to predict tumor purity using miRNAs are still lacking. We predicted tumor purity using miRNA expression data for 16 TCGA tumor types using random forest regression. In addition, we identified miRNAs with high feature-importance scores and examined the extent of the change in predictive performance using informative miRNAs. The predictive performance obtained using only 10 miRNAs with high feature importance was close to the result obtained using all miRNAs. Furthermore, we also found genes targeted by miRNAs and confirmed that these genes were mainly related to immune and cancer pathways. Therefore, we found that the miRNA expression data could predict tumor purity well, and the results suggested the possibility that 10 miRNAs with high feature importance could be used as potential markers to predict tumor purity and to help improve our understanding of the tumor microenvironment.

Prediction of tumor purity from gene expression data using machine learning.

MOTIVATION: Bulk tumor samples used for high-throughput molecular profiling are often an admixture of cancer cells and non-cancerous cells, which include immune and stromal cells. The mixed composition can confound the analysis and affect the biological interpretation of the results, and thus, accurate prediction of tumor purity is critical. Although several methods have been proposed to predict tumor purity using high-throughput molecular data, there has been no comprehensive study on machine learning-based methods for the estimation of tumor purity. RESULTS: We applied various machine learning models to estimate tumor purity. Overall, the models predicted the tumor purity accurately and showed a high correlation with well-established gold standard methods. In addition, we identified a small group of genes and demonstrated that they could predict tumor purity well. Finally, we confirmed that these genes were mainly involved in the immune system. AVAILABILITY: The machine learning models constructed for this study are available at https://github.com/BonilKoo/ML_purity.

Novel deep learning-based survival prediction for oral cancer by analyzing tumor-infiltrating lymphocyte profiles through CIBERSORT.

The tumor microenvironment (TME) within mucosal neoplastic tissue in oral cancer (ORCA) is greatly influenced by tumor-infiltrating lymphocytes (TILs). Here, a clustering method was performed using CIBERSORT profiles of ORCA data that were filtered from the publicly accessible data of patients with head and neck cancer in The Cancer Genome Atlas (TCGA) using hierarchical clustering where patients were regrouped into binary risk groups based on the clustering-measuring scores and survival patterns associated with individual groups. Based on this analysis, clinically reasonable differences were identified in 16 out of 22 TIL fractions between groups. A deep neural network classifier was trained using the TIL fraction patterns. This internally validated classifier was used on another individual ORCA dataset from the International Cancer Genome Consortium data portal, and patient survival patterns were precisely predicted. Seven common differentially expressed genes between the two risk groups were obtained. This new approach confirms the importance of TILs in the TME and provides a direction for the use of a novel deep-learning approach for cancer prognosis.

Episodic Vestibular Syndrome with Hyperventilation-Induced Downbeat Nystagmus.

Hyperventilation-induced downbeat nystagmus (HV-DBN) has been reported in cerebellar disorders and explained by a loss of the inhibitory cerebellar output via a metabolic effect on cerebellar Ca2+ channels. The aim of this study was to determine the clinical characteristics and underlying pathogenesis of episodic vestibular syndrome (EVS) with HV-DBN. Of 667 patients with EVS, we recruited 22 with HV-DBN and assessed their clinical characteristics, video-oculographic findings, and the results of molecular genetic analyses. The age at symptom onset was 47.5 ± 13.0 years (mean ± SD), and there was a female preponderance (n = 15, 68%). The duration of vertigo/dizziness attacks ranged from minutes to a few days, and 11 patients (50%) fulfilled the diagnostic criteria for vestibular migraine. HV-induced new-onset DBN in 8 patients, while the remaining 14 showed augmentation of spontaneous DBN by HV. The maximum slow-phase velocity of HV-DBN ranged from 2.2 to 11.9°/s, which showed a statistical difference with that of spontaneous DBN (median = 4.95, IQR = 3.68-6.55 vs. median = 1.25, IQR = 0.20-2.15, p < 0.001). HV-DBN was either purely downbeat (n = 11) or accompanied with small horizontal components (n = 11). Other neuro-otologic findings included perverted head-shaking nystagmus (n = 11), central positional nystagmus (n = 7), saccadic pursuit (n = 3), and horizontal gaze-evoked nystagmus (n = 1). Gene expression profiling with a bioinformatics analysis identified 43 upregulated and 49 downregulated differentially expressed genes (DEGs) in patients with EVS and HV-DBN and revealed that the downregulated DEGs were significantly enriched in terms related to the ribosome pathway. Our results suggest that the underlying cerebellar dysfunction would be responsible for paroxysmal attacks of vertigo in patients with EVS and HV-DBN.

Neutrophil-mediated immune response as a possible mechanism of acute unilateral vestibulopathy.

OBJECTIVE: This study aimed to investigate the underlying pathogenesis of acute unilateral vestibulopathy (AUV) using gene expression profiling combined with bioinformatics analysis. METHODS: Total RNA was extracted from the peripheral blood mononuclear cells of ten AUV patients in the acute phase and from ten controls. The differentially expressed genes (DEGs) between these two groups were screened using microarray analysis with the cut-off criteria (|fold changes| > 1.5 and p-value < 0.05). Functional enrichment analysis of DEGs was performed using Gene Ontology and Kyoto Encyclopedia of Genes and Genomes pathway analysis, and the protein-protein interaction (PPI) network was constructed using the STRING (Search Tool for the Retrieval of Interacting Genes) database. RESULTS: There were 57 DEGs (50 up-regulated and 7 down-regulated) identified in the AUV group. Functional enrichment analysis showed that most of the up-regulated DEGs were significantly enriched in terms related to the neutrophil-mediated immune pathway. From the PPI network, the top ten hub genes were extracted by calculating four topological properties, and most of them were related to the innate immune system, inflammatory processes and vascular disorders. The complete blood count tests showed that the neutrophil-to-lymphocyte ratio was significantly higher in the 72 AUV patients than in the age-matched controls (2.93±2.25 vs 1.54±0.61, p < 0.001). CONCLUSIONS: This study showed that the neutrophil-mediated immune pathway may contribute to the development of AUV by mediating inflammatory and thrombotic changes in the vestibular organ.

Risk of metastatic pheochromocytoma and paraganglioma in SDHx mutation carriers: a systematic review and updated meta-analysis.

BACKGROUND: Pheochromocytoma and paraganglioma (PPGL) are tumours that arise from chromaffin cells. Some genetic mutations influence PPGL, among which, those in genes encoding subunits of succinate dehydrogenase (SDHA, SDHB, SDHC and SDHD) and assembly factor (SDHAF2) are the most relevant. However, the risk of metastasis posed by these mutations is not reported except for SDHB and SDHD mutations. This study aimed to update the metastatic risks, considering prevalence and incidence of each SDHx mutation, which were dealt formerly all together. METHODS: We searched EMBASE and MEDLINE and selected 27 articles. The patients included in the studies were divided into three groups depending on the presence of PPGL. We checked the heterogeneity between studies and performed a meta-analysis using Hartung-Knapp-Sidik-Jonkman method based on a random effect model. RESULTS: The highest PPGL prevalence was for SDHB mutation, ranging from 23% to 31%, and for SDHC mutation (23%), followed by that for SDHA mutation (16%). The lowest prevalence was for SDHD mutation, ranging from 6% to 8%. SDHAF2 mutation showed no metastatic events. The PPGL incidence showed a tendency similar to that of its prevalence with the highest risk of metastasis posed by SDHB mutation (12%-41%) and the lowest risk by SDHD mutation (~4%). CONCLUSION: There was no integrated evidence of how SDHx mutations are related to metastatic PPGL. However, these findings suggest that SDHA, SDHB and SDHC mutations are highly associated and should be tested as indicators of metastasis in patients with PPGL.

Fast single individual haplotyping method using GPGPU.

BACKGROUND: Most bioinformatic tools for next generation sequencing (NGS) data are computationally intensive, requiring a large amount of computational power for processing and analysis. Here the utility of graphic processing units (GPUs) for NGS data computation is assessed. METHOD: In a previous study, we developed a probabilistic evolutionary algorithm with toggling for haplotyping (PEATH) method based on the estimation of distribution algorithm and toggling heuristic. Here, we parallelized the PEATH method (PEATH/G) using general-purpose computing on GPU (GPGPU). RESULTS: The PEATH/G runs approximately 46.8 times and 25.4 times faster than PEATH on the NA12878 fosmid-sequencing dataset and the HuRef dataset, respectively, with an NVIDIA GeForce GTX 1660Ti. Moreover, the PEATH/G is approximately 13.3 times faster on the fosmid-sequencing dataset, even with an inexpensive conventional GPGPU (NVIDIA GeForce GTX 950). CONCLUSIONS: PEATH/G can be a practical single individual haplotyping tool in terms of both its accuracy and speed. GPGPU can help reduce the running time of NGS analysis tools.

Rare Variants of Putative Candidate Genes Associated With Sporadic Meniere's Disease in East Asian Population.

Objectives: The cause of Meniere's disease (MD) is unclear but likely involves genetic and environmental factors. The aim of this study was to investigate the genetic basis underlying MD by screening putative candidate genes for MD. Methods: Sixty-eight patients who met the diagnostic criteria for MD of the Barany Society were included. We performed targeted gene sequencing using next generation sequencing (NGS) panel composed of 45 MD-associated genes. We identified the rare variants causing non-synonymous amino acid changes, stop codons, and insertions/deletions in the coding regions, and excluded the common variants with minor allele frequency >0.01 in public databases. The pathogenicity of the identified variants was analyzed by various predictive tools and protein structural modeling. Results: The average read depth for the targeted regions was 1446.3-fold, and 99.4% of the targeted regions were covered by 20 or more reads, achieving the high quality of the sequencing. After variant filtering, annotation, and interpretation, we identified a total of 15 rare heterozygous variants in 12 (17.6%) sporadic patients. Among them, four variants were detected in familial MD genes (DTNA, FAM136A, DPT), and the remaining 11 in MD-associated genes (PTPN22, NFKB1, CXCL10, TLR2, MTHFR, SLC44A2, NOS3, NOTCH2). Three patients had the variants in two or more genes. All variants were not detected in our healthy controls (n = 100). No significant differences were observed between patients with and without a genetic variant in terms of sex, mean age of onset, bilaterality, the type of MD, and hearing threshold at diagnosis. Conclusions: Our study identified rare variants of putative candidate genes in some of MD patients. The genes were related to the formation of inner ear structures, the immune-associated process, or systemic hemostasis derangement, suggesting the multiple genetic predispositions in the development of MD.

Identification of Local Clusters of Mutation Hotspots in Cancer-Related Genes and Their Biological Relevance.

Mutation hotspots are either solitary amino acid residues or stretches of amino acids that show elevated mutation frequency in cancer-related genes, but their prevalence and biological relevance are not completely understood. Here, we developed a Smith-Waterman algorithm-based mutation hotspot discovery method, MutClustSW, to identify mutation hotspots of either single or clustered amino acid residues. We identified 181 missense mutation hotspots from COSMIC and TCGA mutation databases. In addition to 77 single amino acid residue hotspots (42.5 percent) including well-known mutation hotspots such as IDH1 (p.R132) and BRAF (p.V600), we identified 104 mutation hotspots (57.5 percent) as clusters or stretches of multiple amino acids, and the hotspots on MUC2, EPPK1, KMT2C, and TP53 were larger than 50 amino acids. Twelve of 27 nonsense mutation hotspots (44.4 percent) were observed in four cancer-related genes, TP53, ARID1A, CDKN2A, and PTEN, suggesting that truncating mutations on some tumor suppressor genes are not randomly distributed as previously assumed. We also show that hotspot mutations have higher mutation allele frequency than non-hotspots, and the hotspot information can be used to prioritize the cancer drivers. Together, the proposed algorithm and the mutation hotspot information can serve as valuable resources in the selection of functional driver mutations and associated genes.

A drug-repositioning screen for primary pancreatic ductal adenocarcinoma cells identifies 6-thioguanine as an effective therapeutic agent for TPMT-low cancer cells.

Pancreatic cancer is one of the most difficult cancers to cure due to the lack of early diagnostic tools and effective therapeutic agents. In this study, we aimed to isolate new bioactive compounds that effectively kill pancreatic ductal adenocarcinoma (PDAC) cells, but not untransformed, human pancreatic ductal epithelial (HPDE) cells. To this end, we established four primary PDAC cell lines and screened 4141 compounds from four bioactive-compound libraries. Initial screening yielded 113 primary hit compounds that caused over a 50% viability reduction in all tested PDAC cells. Subsequent triplicate, dose-dependent analysis revealed three compounds with a tumor cell-specific cytotoxic effect. We found that these three compounds fall into a single category of thiopurine biogenesis. Among them, 6-thioguanine (6-TG) showed an IC50 of 0.39-1.13 µm toward PDAC cells but had no effect on HPDE cells. We propose that this cancer selectivity is due to differences in thiopurine methyltransferase (TPMT) expression between normal and cancer cells. This enzyme is responsible for methylation of thiopurine, which reduces its cytotoxicity. We found that TPMT levels were lower in all four PDAC cell lines than in HPDE or Panc1 cells, and that knockdown of TPMT in HPDE or Panc1 cells sensitized them to 6-TG. Lastly, we used a patient-derived xenograft model to confirm that 6-TG has a significant antitumor effect in combination with gemcitabine. Overall, our study presents 6-TG as a strong candidate for use as a therapeutic agent against PDAC with low levels of TPMT.

Integrative Bioinformatics and Functional Analyses of GEO, ENCODE, and TCGA Reveal FADD as a Direct Target of the Tumor Suppressor BRCA1.

BRCA1 is a multifunctional tumor suppressor involved in several essential cellular processes. Although many of these functions are driven by or related to its transcriptional/epigenetic regulator activity, there has been no genome-wide study to reveal the transcriptional/epigenetic targets of BRCA1. Therefore, we conducted a comprehensive analysis of genomics/transcriptomics data to identify novel BRCA1 target genes. We first analyzed ENCODE data with BRCA1 chromatin immunoprecipitation (ChIP)-sequencing results and identified a set of genes with a promoter occupied by BRCA1. We collected 3085 loci with a BRCA1 ChIP signal from four cell lines and calculated the distance between the loci and the nearest gene transcription start site (TSS). Overall, 66.5% of the BRCA1-bound loci fell into a 2-kb region around the TSS, suggesting a role in transcriptional regulation. We selected 45 candidate genes based on gene expression correlation data, obtained from two GEO (Gene Expression Omnibus) datasets and TCGA data of human breast cancer, compared to BRCA1 expression levels. Among them, we further tested three genes (MEIS2, CKS1B and FADD) and verified FADD as a novel direct target of BRCA1 by ChIP, RT-PCR, and a luciferase reporter assay. Collectively, our data demonstrate genome-wide transcriptional regulation by BRCA1 and suggest target genes as biomarker candidates for BRCA1-associated breast cancer.

Population-based statistical inference for temporal sequence of somatic mutations in cancer genomes.

BACKGROUND: It is well recognized that accumulation of somatic mutations in cancer genomes plays a role in carcinogenesis; however, the temporal sequence and evolutionary relationship of somatic mutations remain largely unknown. METHODS: In this study, we built a population-based statistical framework to infer the temporal sequence of acquisition of somatic mutations. Using the model, we analyzed the mutation profiles of 1954 tumor specimens across eight tumor types. RESULTS: As a result, we identified tumor type-specific directed networks composed of 2-15 cancer-related genes (nodes) and their mutational orders (edges). The most common ancestors identified in pairwise comparison of somatic mutations were TP53 mutations in breast, head/neck, and lung cancers. The known relationship of KRAS to TP53 mutations in colorectal cancers was identified, as well as potential ancestors of TP53 mutation such as NOTCH1, EGFR, and PTEN mutations in head/neck, lung and endometrial cancers, respectively. We also identified apoptosis-related genes enriched with ancestor mutations in lung cancers and a relationship between APC hotspot mutations and TP53 mutations in colorectal cancers. CONCLUSION: While evolutionary analysis of cancers has focused on clonal versus subclonal mutations identified in individual genomes, our analysis aims to further discriminate ancestor versus descendant mutations in population-scale mutation profiles that may help select cancer drivers with clinical relevance.

PEATH: single-individual haplotyping by a probabilistic evolutionary algorithm with toggling.

Motivation: Single-individual haplotyping (SIH) is critical in genomic association studies and genetic diseases analysis. However, most genomic analysis studies do not perform haplotype-phasing analysis due to its complexity. Several computational methods have been developed to solve the SIH problem, but these approaches have not generated sufficiently reliable haplotypes. Results: Here, we propose a novel SIH algorithm, called PEATH (Probabilistic Evolutionary Algorithm with Toggling for Haplotyping), to achieve more accurate and reliable haplotyping. The proposed PEATH method was compared to the most recent algorithms in terms of the phased length, N50 length, switch error rate and minimum error correction. The PEATH algorithm consistently provides the best phase and N50 lengths, as long as possible, given datasets. In addition, verification of the simulation data demonstrated that the PEATH method outperforms other methods on high noisy data. Additionally, the experimental results of a real dataset confirmed that the PEATH method achieved comparable or better accuracy. Availability and implementation: Source code of PEATH is available at https://github.com/jcna99/PEATH. Contact: jkrhee@catholic.ac.kr or sooyong.shin@gmail.com. Supplementary information: Supplementary data are available at Bioinformatics online.

Impact of Tumor Purity on Immune Gene Expression and Clustering Analyses across Multiple Cancer Types.

Surgical archives of tumor specimens are often impure. The presence of RNA transcripts from nontumor cells, such as immune and stromal cells, can impede analyses of cancer expression profiles. To systematically analyze the impact of tumor purity, the gene expression profiles and tumor purities were obtained for 7,794 tumor specimens across 21 tumor types (available in The Cancer Genome Atlas consortium). First, we observed that genes with roles in immunity and oxidative phosphorylation were significantly inversely correlated and correlated with the tumor purity, respectively. The expression of genes implicated in immunotherapy and specific immune cell genes, along with the abundance of immune cell infiltrates, was substantially inversely correlated with tumor purity. This relationship may explain the correlation between immune gene expression and mutation burden, highlighting the need to account for tumor purity in the evaluation of expression markers obtained from bulk tumor transcriptome data. Second, examination of cluster membership of gene pairs, with or without controlling for tumor purity, revealed that tumor purity may have a substantial impact on gene clustering across tumor types. Third, feature genes for molecular taxonomy were analyzed for correlation with tumor purity, and for some tumor types, feature genes representing the mesenchymal and classical subtypes were inversely correlated and correlated with tumor purity, respectively. Our findings indicate that tumor purity is an important confounder in evaluating the correlation between gene expression and clinicopathologic features such as mutation burden, as well as gene clustering and molecular taxonomy. Cancer Immunol Res; 6(1); 87-97. ©2017 AACR.

The association between the nicotinic acetylcholine receptor α4 subunit gene (CHRNA4) rs1044396 and Internet gaming disorder in Korean male adults.

The primary aim of this study was to investigate the genetic predisposition of Internet gaming disorder (IGD), and the secondary aim was to compare the results to those of alcohol dependence (AD). Two independent case-control studies were conducted. A total of 30 male participants with IGD, diagnosed according to the 5th edition of the Diagnostic and Statistical Manual of Mental Disorders (DSM-5) criteria, and 30 sex-matched controls participated in study 1. We designed targeted exome sequencing (TES) to test for 72 candidate genes that have been implicated in the pathogenesis of addiction. The genes included seven neurotransmitter (dopamine, serotonin, glutamate, r-aminobutyric acid (GABA), norepinephrine, acetylcholine, and opioid) system genes. A total of 31 male in-patients with AD and 29 normal male controls (NC) were enrolled in study 2. The same 72 genes included in study 1 and ten additional genes related to alcohol-metabolic enzyme were selected as the target genes, and we identified the genetic variants using the same method (TES). The IGD group had a lower frequency of the T allele of rs1044396 in the nicotinic acetylcholine receptor alpha 4 subunit (CHRNA4), and this variant represents a protective allele against IGD. However, we did not find a significant difference in the polymorphisms of the 72 genes that encode neurotransmitter systems between the AD and NC groups. This study demonstrated that rs1044396 of CHRNA4 was significantly associated with IGD.

Passage-dependent accumulation of somatic mutations in mesenchymal stromal cells during in vitro culture revealed by whole genome sequencing.

Human mesenchymal stromal cells (MSCs) have served as a major cellular resource for cell-based immunomodulatory and regenerative therapies. However, genomic instability may accumulate during ex vivo expansion of MSCs, thereby increasing the potential of malignant transformation. Here, we performed whole genome sequencing of two peripheral blood-derived MSC lines (MSC1 and MSC2) at various passages (passage 1 [P1] to P9). The majority of single-nucleotide variations (SNVs) occurred in later passages; specifically, 90% and 70% of all SNVs in MSC1 and MSC2 were observed in P9 and P7/P9, respectively. These late-occurring SNVs were enriched with C > A transversions and were overrepresented in intronic regions compared to intergenic regions, suggesting that the mutational forces are not constant across the passages. Clonality analyses also distinguished early-occurring, subclonal SNVs from late-occurring, clonally fixed SNVs. In addition, MSCs were largely devoid of copy number alterations (CNAs) (i.e., 0-2 CNAs per passage), with one exception (MSC2-P3) harboring 29 passage-specific CNAs. Our findings suggest that the SNVs found to be abundant at later passages likely resulted from the accumulation of replication stress, which can be associated with proliferation activity. Thus, the genomic instability associated with proliferation records should be considered for clinical applications of MSCs.

Allelic imbalance of somatic mutations in cancer genomes and transcriptomes.

Somatic mutations in cancer genomes often show allelic imbalance (AI) of mutation abundance between the genome and transcriptome, but there is not yet a systematic understanding of AI. In this study, we performed large-scale DNA and RNA AI analyses of >100,000 somatic mutations in >2,000 cancer specimens across five tumor types using the exome and transcriptome sequencing data of the Cancer Genome Atlas consortium. First, AI analysis of nonsense mutations and frameshift indels revealed that nonsense-mediated decay is typical in cancer genomes, and we identified the relationship between the extent of AI and the location of mutations in addition to the well-recognized 50-nt rules. Second, the AI with splice site mutations may reflect the extent of intron retention and is frequently observed in known tumor suppressor genes. For missense mutations, we observed that mutations frequently subject to AI are enriched to genes related to cancer, especially those of apoptosis and the extracellular matrix, and C:G > A:T transversions. Our results suggest that mutations in known cancer-related genes and their transcripts are subjected to different levels of transcriptional or posttranscriptional regulation compared to wildtype alleles and may add an additional regulatory layer to the functions of cancer-relevant genes.