Cases Overview
Clinical information
Mutations
Copy Number Alterations
Follow the three steps to use this tool:
Preparation: download the template and fill the form.
Imputation: upload the filled form and click “Imputed data” to view the data.
Prediction: click “Predicted result” to view the predicted HM status.
Load example
Download the template Template
Download the feature description Description
Select the read.table parameters below
Download the template Template
Download the feature description Description
Download data
Below we provide the raw data that we published in the STM2023 paper for research use.
- Clinical information of the cohort: Click here
- Somatic mutations of the glioma samples: Click here
- mRNA expression of the initial gliomas: Click here
- mRNA expression of the recurrent gliomas: Click here
- Updated grading (WHO 2021) of the TCGA LGG cohort: Click here
Citation:
If you use these data in your publication, please consider cite:
Introduction
CELLO2 is an interactive web server to visualize curated clinical and somatic genomic information of longitudinally paired gliomas, and to explore the probability of developing treatment-induced hypermutation using models that we trained from the collected data. The app is hosted on shinyapps.io at http://www.wang-lab-hkust.com:3838/cello2.
The server is built in R and Shiny, and is right now under active updating. In this web site, you can:
- Click Genes panel to explore genomic alterations and RNA expression of one specific gene;
- Click Cases panel to explore clinical and genetic information of patients;
- Click CELLO2 panel to explore the probability of developing treatment-induced hypermutation using models that we trained from the collected data;
In this page you can visualize the genomic alteration and RNA expression of one specific gene in the initial and recurrent tumors of glioma patients. It is also possible to link the changes in genomic alteration and/or gene expression with several important clinical features such as molecular subtype, pathological grade, grade progression and development of hypermutation.
Visualize genomic alteration and gene expression
In the left control panel, you can input the gene name, such as IDH1, then click submit. At the right side of the page, you will see two figures displaying the gene expression (upper panel) and genomic alteration (lower panel) in the cohort. By default, the figures are separated into two facets representing hypermutation status at recurrence. Select to group by Molecular_Subtype and we can see the gene expression data are shown in scatter plot, and each point represents the expression level of one sample, in the unit of log2 RPKM. The genomic alteration data are shown in a heatmap. It is clear to see that IDH1 mutation are present in all IDH-mutant-codel and IDH-mutant-non-codel patients but in none of the IDHwt patients:
Regarding genomic alteration, we included data for mutations, copy number alterations and several other important alterations such as FGFR3-TACC3 fusion, PTPRZ1-MET fusion, EGFR*vIII, *MET*ex14, *MGMT fusions and hypermutation (HM). For example, to check hypermutated samples in this cohort, you can simply type HM in the gene name box.
Group by clinical features for comparison
In addition to molecular subtype, we can also explore other clinical features. For example, enter MKI67 in the gene name box, and select to group by Hypermutation_at_recurrence, from the results we can see higher frequency of MKI67 overexpression in patients that develop hypermutation:
An additional useful feature is to show the longitudinal pairs in the gene expression plot. This will add a link between the initial and recurrent tumor of each patient. For example, if we check VEGFA expression and Grade_Progression, we will see significant elevation of VEGFA expression during grade progression:
In the cases panel, there is an overview of the cases, which shows the time point of each surgery as well as the overall survival. In the input box, one can input a patient ID to view the clinical data, somatic mutations, and copy number alterations.
The “CELLO2” panel performs imputation and prediction of hypermutation (HM) using clinical and genomic profiles.
“example” patients prediction
To make it easier to understand the process of imputation and prediction of hypermutation (HM), we create “example” button which enables users to access the predicted results of three patients.
Click the “example” button.
The “example” patients will be displayed in “Upload data”. The “NA” (not available data) will be highlighted by yellow.
NAs will be imputed by ISEM (iteratively sequential ensemble machine) and users need to click the “Imputed data” panel. The imputation data is colored by red.
After imputation, users can click the “Predicted result” panel to acquire the likelihood of harbouring HM (the left digital dashboard), ranging from 0 (non-HM) to 1 (HM).
Users can access the results of one specific patient by selecting the patient ID from “Choose ID of sample”.
The predicted results of all example patients can be download by clicking the “download” button on the bottom.
“upload” patients prediction
Users can explore the HM and Grade status of patients in their own cohort.
Prepare the data:
Download the template (click the “Template” button).
Fill in the corresponding information according to the 46 features (the description can be downloaded by clicking the “Description” button). For unknown or unavailable features, users can leave blank without filling in the numerical value.
Note: Each row means each patient, and CELLO 2 supports submission of multiple samples implemented by multiple rows.
Upload the data: the csv and txt file can be uploaded.
The upload data can be seen in “Upload data” panel. The “NA” (not available data) will be highlighted by yellow.
NAs will be imputed by ISEM (iteratively sequential ensemble machine) and users need to click the “Imputed data” panel. The imputation data is colored by red.
The probability of harbouring HM ranging from 0 (non-HM) to 1 (HM).
CELLO2 supports batch download, which can enable users to freely get all the predicted results by clicking “download” button on the bottom.
CELLO2: Cancer EvoLution from machine learning of LOngitudinal sequencing
CELLO2 is developed by Yingxi Yang, Quanhua Mu, and Zhihan Zhu in Wang-Lab@HKUST. It further expands the functionality of our previous CELLO toolkit*.
Correspondence: jgwang@ust.hk
Abstract
Clonal evolution drives cancer progression and therapeutic resistance. Recent studies revealed divergent longitudinal trajectories in gliomas but early molecular traits that steer post-treatment cancer evolution remain unclear. We comprehensively analyzed sequencing data of 544 initial-recurrent adult diffuse glioma pairs to identify genomic and transcriptomic early predictors of tumor evolution in each molecular subtype and developed machine-learning methods capable of predicting hypermutation. We validated the association between the top predictor, c-MYC gain / MYC targets activation, and hypermutagenesis by TMZ resistance induction experiments in glioma cell lines and an isogenic model manipulating patient-derived gliomaspheres. We demonstrated that c-Myc, binding to open chromatin and transcriptionally active genomic regions, increases the vulnerability of key mismatch repair genes to TMZ-induced mutagenesis, thus triggering hypermutation. This study reveals early predictors of cancer evolution under therapy and provides rich resources for precision oncology targeting cancer dynamics.
How to cite
Mu et al. Identifying predictors of glioma evolution from longitudinal sequencing. Sci Transl Med. 15.716 (2023): eadh4181.
Sources of the published datasets
^: not including GLASS samples that are overlapping with other sources.
*: Jiang B., Song D., Mu Q., & Wang J. (2020). CELLO: a longitudinal data analysis toolbox untangling cancer evolution. Quantitative Biology, 8(3), 256-266.