AS-CMC has a user-friendly interface, which allows researchers to explore alternative splicing (AS) events in The Cancer Genomics Atlas (TCGA) molecular subtypes. Our web service consists of two parts, viz. “Single-cancer AS” and “Pan-cancer AS.” In the “Single-cancer AS,” users can select cancer type first and get the list of AS events with the statistical analysis results. In the “Pan-cancer AS,” users can obtain pan-cancer views for a selected AS event.
AS-CMC provides three analysis modules. In the “Subtype-specific AS” module (top left), differential regulation of AS PSI values was tested among molecular subtypes provided by TCGAbiolinks. In the “Phenotype association” module (bottom left), each AS event was evaluated in association with patient-level (clinical outcomes), tissue-level (microenvironment), and gene-level (gene-expression) data. In the “Pan-cancer comparison” module (right), the analyzed data pertaining to each AS event is displayed in a panoramic view across cancer types.
The number of samples in each cancer type is shown in the parenthesis.
(a) Number of patient samples for each cancer. Each bar consists of the molecular subtypes. Subtype names can be displayed above the graph.
(b) The number of subtype-specific AS events by splice types. The subtype-specific AS events were selected based on analysis of variance (ANOVA) (p < 0.001 and adjusted R2 > 0.1).
(c) The fraction of survival-associated AS events among subtype-specific AS events. The fraction is marked by dark red color and is also shown as percentage on the right side of each bar. X-axis indicates the number of AS events.
 
|
 
|
AS-CMC (Alternative Splicing for Cancer Molecular Classification) is a web-based database for allowing users to browse subtype-specific changes in AS along with phenotypic association for each cancer type as well as compare the regulation pattern across diverse cancer types. For AS in TCGA samples, we used the per-cent-spliced-in index (PSI) value from TCGASpliceSeq database. We obtained the information pertaining to cancer molecular subtypes from TCGAbiolinks R package.
For AS of TCGA samples, we download the PSI value of 27,682 AS events in 24 cancer types from the TCGASpliceSeq database. AS-CMC provided the PSI values of five splicing types: exon skip (ES), retained intron (RI), alternate acceptor sites (AA), alternate donor sites (AD), and mutually exclusive exon (ME).
Percent splice in (PSI) ranged from 0 to 1 was a commonly used ratio to indicate different uses of alternative exon. For each splice event, a percent-splice-in (PSI) value generated by the ratio of inclusion of reads over the total reads for that event (both inclusion and exclusion reads).
AS ID rule is simple to create from gene model. AS ID format includes a AS type, exon number and delimiting character '_'.
(AS ID : Gene.Name_AS.Type_Retulated.Exon_From.Exon_To.Exon)
The information of cancer molecular sub-types was obtained from TCGAbiolinks R package. And we used the most prominent subtype classification which is curated annotation named ‘Subtype_Selected’. For more information on the molecular types of each cancer, see the literature links in the table below.
Cancer | Molecular data | Reference | #Type | Molecular Subtype (n) |
---|---|---|---|---|
ACC | DNAmeth | Cancer Cell 2016 | 3 | ACC.CIMP-high, ACC.CIMP-intermediate, ACC.CIMP-low |
BLCA | mRNA | Nature 2014 | 4 | BLCA.4, BLCA.3, BLCA.1, BLCA.2 |
BRCA | PAM50(mRNA) | Nature 2012 | 5 | BRCA.Normal, BRCA.Her2, BRCA.Basal, BRCA.LumB, BRCA.LumA |
COAD | Molecular_Subtype | Cancer Cell 2018 | 4 | GI.HM-SNV, GI.GS, GI.MSI, GI.CIN |
ESCA | Molecular_Subtype | Cancer Cell 2018 | 5 | GI.GS, GI.HM-SNV, GI.MSI, GI.CIN, GI.ESCC |
GBM | Supervised_DNAmeth | Cell 2016 | 6 | GBM_LGG.G-CIMP-high, GBM_LGG.G-CIMP-low, GBM_LGG.LGm6-GBM, GBM_LGG.NA, GBM_LGG.Classic-like, GBM_LGG.Mesenchymal-like |
HNSC | mRNA | Nature 2015 | 4 | HNSC.Classical, HNSC.Atypical, HNSC.Mesenchymal, HNSC.Basal |
KICH | Eosinophilic | Cancer Cell 2014 | 2 | KICH.Eosin.1, KICH.Eosin.0 |
KIRC | mRNA | Nature 2013 | 5 | KIRC.NA, KIRC.4, KIRC.2, KIRC.3, KIRC.1 |
KIRP | COC | NEJM 2015 | 4 | KIRP.C2c - CIMP, KIRP.C2b, KIRP.C2a, KIRP.C1 |
LAML | mRNA | NEJM 2013 | 7 | AML.1, AML.3, AML.7, AML.2, AML.5, AML.6, AML.4 |
LGG | Supervised_DNAmeth | Cell 2016 | 7 | GBM_LGG.NA, GBM_LGG.G-CIMP-low, GBM_LGG.Classic-like, GBM_LGG.PA-like, GBM_LGG.Mesenchymal-like, GBM_LGG.Codel, GBM_LGG.G-CIMP-high |
LIHC | iCluster | Cell 2017 | 4 | LIHC.NA, LIHC.iCluster:2, LIHC.iCluster:3, LIHC.iCluster:1 |
LUAD | iCluster | Nature 2014 | 6 | LUAD.1, LUAD.2, LUAD.4, LUAD.6, LUAD.3, LUAD.5 |
LUSC | mRNA | Nature 2012 | 4 | LUSC.primitive, LUSC.basal, LUSC.secretory, LUSC.classical |
OV | mRNA | Nature 2011 | 4 | OVCA.Immunoreactive, OVCA.Mesenchymal, OVCA.Differentiated, OVCA.Proliferative |
PCPG | mRNA | Cancer Cell 2017 | 5 | PCPG.NA, PCPG.Cortical admixture, PCPG.Wnt-altered, PCPG.Pseudohypoxia, PCPG.Kinase signaling |
PRAD | Mutation/Fusion | Cell 2015 | 8 | PRAD.7-IDH1, PRAD.4-FLI1, PRAD.6-FOXA1, PRAD.3-ETV4, PRAD.2-ETV1, PRAD.5-SPOP, PRAD.8-other, PRAD.1-ERG |
READ | Molecular_Subtype | Cancer Cell 2018 | 4 | GI.MSI, GI.HM-SNV, GI.GS, GI.CIN |
SKCM | Mutation | Cell 2015 | 5 | SKCM.NF1_Any_Mutants, SKCM.Triple_WT, SKCM.- , SKCM.RAS_Hotspot_Mutants, SKCM.BRAF_Hotspot_Mutants |
STAD | Molecular_Subtype | Cancer Cell 2018 | 5 | GI.HM-SNV, GI.EBV, GI.GS, GI.MSI, GI.CIN |
THCA | mRNA | Cell 2014 | 6 | THCA.NA, THCA.2, THCA.3, THCA.5, THCA.4, THCA.1 |
UCEC | iCluster - updated according to Pan-Gyne/Pathways groups | Nature 2013 | 5 | UCEC.NA, UCEC.POLE, UCEC.MSI, UCEC.CN_LOW, UCEC.CN_HIGH |
UCS | mRNA | Cancer Cell 2017 | #2 | UCS.1, UCS.2 |
We performed ANOVA test to selected subtype-specific AS events by analysis of variance (p-value < 1x10 -3) for each cancer type. For visualization,AS-CMC displays boxplots for a each cancer and a scatter plot across cancers.
The “Phenotype association” module enables prioritization of AS events by relevance in terms of clinical outcomes (patient-level), cancer microenvironment scores (tissue-level), and gene expression levels (gene-level).
In this section show two workflows(Tissue-level & Gene-level) of a module. Tissue-level: In the cancer microenvironment part,users can investigate the correlations between the changes in AS and predefined scores indicating the status of immune and stromal cells, epithelial-to-mesenchymal transition (EMT), and hypoxia. Gene-level: Users can also examine the correlations of each AS event with expression levels of all genes. The gene expression levels correlated with AS PSI values enables determination of biological pathways underlying AS changes. The relationship of the host gene with AS helps in assessing the dependency of the AS event on transcriptional regulation. If the AS is independent of gene expression, it is likely regulated solely by splicing machinery
This page shows the approches for the cancer. Method for searching a significant molecular subtype-specific AS events in certain cancer is as follows.
a. List of subtype-specific AS events for a selected cancer type. Once a cancer type is selected, the subtype-specific AS events are shown with relevant statistics displayed in a tabular form. Users can filter the results using the defined cut-off in survival and correlation with the expression level of genes with the AS. b.Visualization panel showing the analysis results of an AS event. Once an AS event is selected (click the hyper linked letter named "Click"), a window with various plots pops up.
We performed survival analysis between high PSI and low PSI cohorts. Threshold PSI value to separate two groups was defined as 10%, 25%, and 50% of PSI value in the distribution. The significance for differential survival rates was evaluated by Log-rank test. Three thresholds in PSI value distribution were applied to generate two groups to be tested. P-value derived from Log-rank test. A pan-cancer view of survival analysis results for the queried AS events.
We compared exon-level PSI values and gene expression levels across cancer molecular subtypes. And we checked the correlation of PSI value to spliced gene expression across cancer molecular subtypes.
To identify potential clinically association of a significant AS event, we analysis survival analysis using TCGA survival data. .
We performed survival analysis between high PSI and low PSI cohorts. Threshold PSI value to separate two groups was defined as 10%, 25%, and 50% of PSI value in the distribution. The significance for differential survival rates was evaluated by Log-rank test. Three thresholds in PSI value distribution were applied to generate two groups to be tested. P-value derived from Log-rank test. A pan-cancer view of survival analysis results for the queried AS events.
To test the relationship of each AS event with tumor microenvironment, we used immune infiltration levels, tumor purity,hypoxia and EMT, which were previously published in pan-cancer analysis. First, we checked if PSI values are related to tumor microenvironment using cellular fraction estimates (leukocyte fraction and CIBERSORT immune fractions) [1]. Next, we assesed the association of PSI values with tumor hypoxia scores which can reflect the level of molecular oxygen in tumor sample [2]. Finally, we provided the association of EMT value [3].
We explored which genes have mRNA levels correlated with queried AS.
This page shows the approches for the cancer.Method for searching a significant molecular subtype-specific AS events in certain cancer is as follows.
a. Selection panel for AS events. Users are directed to the pan-cancer views upon clicking (click the hyper linked letter named "Click") on an AS event. b-d. Pan-cancer information pertaining to MAP3K7 exon 11 as an example. b. Comparison of subtype-specificity across cancer types. ANOVA results are displayed using two different y-axes: -log10.
We performed ANOVA test to selected subtype-specific AS events by analysis of for each cancer type. And then we identify the distribuion p & r-value across cancers
We performed survival analysis between high PSI and low PSI cohorts. Threshold PSI value to separate two groups was defined as 10%, 25%, and 50% of PSI value in the distribution. The significance for differential survival rates was evaluated by Log-rank test. Three thresholds in PSI value distribution were applied to generate two groups to be tested. P-value derived from Log-rank test. A pan-cancer view of survival analysis results for the queried AS events.
Pathways | Genes |
---|---|
Cell Cycle Control (3 genes) | CDK2, E2F5, E2F6 |
Notch signaling (11 genes) | APH1A, ARRDC1, DLL3, DTX2, DTX3, HES4, ITCH, JAG2, NCOR2, NUMB, RBPJ |
DNA Damage Response (4 genes) | CHEK1, CHEK2, RAD51, MLH1 |
Other growth/proliferation signaling (4 genes) | FGFR1, CSF1, PLAGL1, AURKA |
Survival/cell death regulation signaling (4 genes) | BCL2L1, BCL2, CASP10, CASP3 |
RTK signaling family (3 genes) | FGFR1, VEGFA, PDGFA |
PI3K-AKT-mTOR signaling (4 genes) | TSC2, MLST8, AKT1, RHEB |
Ras-Raf-MEK-Erk/JNK signaling (8 genes) | MAPK9, HRAS, MAP3K4, KRAS, MAPK3, MAP3K3, DAB2, MAPK8 |
Regulation of ribosomal protein synthesis and cell growth (3 genes) | RPS6, RPS6KB2, RPS6KB1 |
Angiogenesis (1 gene) | VEGFA |
Invasion nad metastatsis (4 genes) | MMP23B, PTK2, MMP19, WFDC2 |
TGF-beta Pathway (2 genes) | TGFBR3, SMAD5 |
Using AS-CMC, we selected a notable subtype-specific AS event, which can serve as a pan-cancer AS biomarker as an example. An ES event in exon 11 of MAP3K7 (mitogen-activated protein kinase kinase kinase 7) gene was chosen as this marker showed significant subtype-specificity in 10 cancer types (BRCA, LGG, ESCA, STAD, HNSC, BLCA, OV, THCA, LUAD, UCS). In the survival analysis, the skipping of the exon was found to be strongly associated with poor clinical outcome in stomach adenocarcinoma (STAD). The survival difference was found to be the largest in the most stringent cut-off (upper 10% vs lower 10%).
MAP3K7 AS was analyzed in depth in STAD due to its significant association with survival. Among the five molecular subtypes of STAD, only GI.GS subtype showed significant distribution of PSI values compared to the other subtypes. The PSI values were correlated with the molecular scores related to EMT scores (r = -0.71). Taken together, these data support that an ES event in exon 11 of MAP3K7 may play a role in regulating subtypes across diverse cancers, and that this event may particularly play a crucial role in STAD where its function has not been reported earlier.
An example of a potential pan-cancer AS biomarker. (a) Location of MAP3K7 exon with AS on the corresponding gene and chromosome. (b) Distribution of PSI values for five molecular subtypes in STAD. (c) Correlations between the AS PSIs and EMT scores in STAD. Spearman’s correlation coefficient is shown on the top. Each dot represents an individual patient sample. (d) Survival plots comparing survival rates between two patient groups with high- and low-PSI values in STAD. In the survival plots, AS-CMC provides three plots representing the survival difference between the groups for an AS event according to the following PSI cut-offs: 50% (upper 50% vs lower 50%), 25% (upper 25% vs lower 25%), and 10% (upper 10% vs lower 10%). The significance of differential survival rates was evaluated using log rank test.
If you have any questions or suggestions on this database, please feel free to contact us.
Jiyeon Park PhD
E.mail : parkji7@gmail.com
Precision Medicine Research Center, College of Medicine, The Catholic University of Korea, Seoul, Republic of Korea
Integrated Research Center for Genome Polymorphism, College of Medicine, The Catholic University of Korea, Seoul, Republic of Korea
Jin-Ok Lee MSc
E.mail : jinoklee.01@gmail.com
Precision Medicine Research Center, College of Medicine, The Catholic University of Korea, Seoul, Republic of Korea
Integrated Research Center for Genome Polymorphism, College of Medicine, The Catholic University of Korea, Seoul, Republic of Korea
Yeun-Jun Chung MD, PhD
E.mail : yejun@catholic.ac.kr
Precision Medicine Research Center, College of Medicine, The Catholic University of Korea, Seoul, Republic of Korea
Integrated Research Center for Genome Polymorphism, College of Medicine, The Catholic University of Korea, Seoul, Republic of Korea
Department of Microbiology, College of Medicine, The Catholic University of Korea, Seoul, Republic of Korea