Machine Learning–Based Radiomics for Molecular Subtyping of Gliomas

Lu, Chia‐Feng; Hsu, Fei Ting; Hsieh, Kevin; Kao, Yu Chieh Jill; Cheng, Shih‐Ping; Hsu, Justin Bo Kai; Tsai, Pei‐Hsuan; Chen, Ray Jade; Chao, Huang; Yen, Yun; Chen, Cheng-Yu

doi:10.1158/1078-0432.ccr-17-3445

Cited by 237 publications

(176 citation statements)

References 33 publications

Supporting

Mentioning

171

Contrasting

Unclassified

Order By: Relevance

“…The 2016 WHO classification of central nervous system (CNS) tumors combines molecular parameters and histology to define diffuse gliomas . Based on traditional histopathology but enriched with IDH and 1p/19q codeletion status, gliomas could be classified into five subtypes (three LGG and two GBM), as follows: (i) LGG with wild‐type IDH (LGG‐IDHwt); (ii) LGG with IDH mutation and 1p/19q non‐codeletion (LGG‐IDHmut‐noncodel); (iii) LGG with IDH mutation and 1p/19q codeletion (LGG‐IDHmut‐codel); (iv) GBM with wild‐type IDH (GBM‐IDHwt); and (v) GBM with IDH mutation (GBM‐IDHmut) . These five subtypes of glioma show distinct tumor characteristics and OS outcomes.…”

Section: Introductionmentioning

confidence: 99%

Amino acid metabolism‐related gene expression‐based risk signature can better predict overall survival for glioma

et al. 2018

View full text Add to dashboard Cite

Metabolic reprogramming has been proposed to be a hallmark of cancer. Aside from the glycolytic pathway, the metabolic changes of cancer cells primarily involve amino acid metabolism. However, in glioma, the characteristics of the amino acid metabolism‐related gene set have not been systematically profiled. In the present study, RNA sequencing expression data from 309 patients in the Chinese Glioma Genome Atlas database were included as a training set, while another 550 patients within The Cancer Genome Atlas database were used to validate. Consensus clustering of the 309 samples yielded two robust groups. Compared with Cluster1, Cluster2 correlated with a better clinical outcome. We then developed an amino acid metabolism‐related risk signature for glioma. Our results showed that patients in the high‐risk group had dramatically shorter overall survival than low‐risk counterparts in any subgroup, stratified by isocitrate dehydrogenase and 1p/19q status based on the 2016 World Health Organization classification guidelines. The 30‐gene signature showed better prognostic value than the traditional factors “age” and “grade” by analyzing the receiver operating characteristic curve with areas under curve of 0.966, 0.692, 0.898 and 0.975, 0.677, 0.885 for 3‐ and 5‐year survival, respectively. Moreover, univariate and multivariate analysis showed that the 30‐gene signature was an independent prognostic factor for glioma. Furthermore, Gene Ontology analysis and Gene Set Enrichment Analysis showed that tumors with a high risk score correlated with various aspects of the malignancy of glioma. In summary, we demonstrated a novel amino acid metabolism‐related risk signature for predicting prognosis for glioma.

show abstract

Section: Introductionmentioning

confidence: 99%

Amino acid metabolism‐related gene expression‐based risk signature can better predict overall survival for glioma

et al. 2018

View full text Add to dashboard Cite

show abstract

“…With the availability of public databases, the integration of data on histology and mutation, methylation and mutation or mRNA expression and mutation can divide patients into different subgroups [4,5,7,27]. Furthermore, with the development of arti cial intelligence and machine learning, digital images obtained via magnetic resonance imaging and histopathological analysis can be used to predict not only overall survival but also IDH mutation and 1p/19q codeletion [18,21]. In the near future, the diagnosis of gliomas will involve the combination of multidimensional data.…”

Section: Discussionmentioning

confidence: 99%

A nuclear transport-related gene signature combined with IDH and 1p/19q better predicts the prognosis of glioma patients

Zhu

Lan

Wang

et al. 2020

Preprint

View full text Add to dashboard Cite

Background The nuclear transport system has been proposed to be indispensable for cell proliferation and invasion in cancers. Prognostic biomarkers and molecular targets in nuclear transport systems have been developed. However, no systematic analysis of genes related to nuclear transport in gliomas has been performed. An integrated prognostic classification involving mutation and nuclear transport gene signatures has not yet been explored. Methods In the present study, we analyzed gliomas from a training cohort (TCGA dataset, n = 660) and validation cohort (CGGA dataset, n = 668) to develop a prognostic nuclear transport gene signature and generate an integrated classification system. Then, we developed a nuclear transport-related risk score (NTRS) for gliomas with a training cohort. Results Gene set enrichment analysis (GSEA) showed that glioblastoma (GBM) was mainly enriched in nuclear transport progress compared to lower-grade glioma (LGG). NTRS was significantly correlated with clinical and genetic characteristics, including grade, age, histology, IDH status and 1p/19q codeletion, in the training and validation cohorts. Survival analysis revealed that patients with a higher NTRS exhibited shorter overall survival. NTRS showed better prognostic value compared to classical molecular markers, including IDH status and 1p/19q codeletion. Furthermore, univariate and multivariate analyses indicated that NTRS was an independent prognostic factor for gliomas. Enrichment map and Gene Ontology analysis demonstrated that signaling pathways related to the cell cycle were enriched in the NTRSHigh group. Subgroup survival analysis revealed that NTRS could differentiate the outcomes of low- and high-risk patients with wild-type IDH or mutant IDH and 1p/19q noncodeletion. Conclusions NTRS is associated with poor outcomes and could be an independent prognostic marker in diffuse gliomas. Prognostic classification combined with IDH mutation, 1p/19q codeletion and NTRS could better predict the survival of glioma patients.

show abstract

“…For GBM subtype classification: Ryu et al [47] proposed a three-level machine-learning model and obtained 0.83 AUC and 0.8 ACC. Lu et al [21] obtained an AUC and ACC of 0.92 and 0.88,…”

Section: Comparison With the State Of The Artmentioning

confidence: 98%

“…Methods Result AUC ACC Pre BRCA Liao [22] N/A 0.87 N/A Guo [20] N/A N/A 0.88 Telonis [41] N/A 0.91 N/A Li [42] 0.89 N/A N/A Sherafatian [43] N/A 0.89 0.90 MLW-gcForest 0.98 0.91 0.92 LUAD Liao [22] N/A 0.91 N/A Guo [20] N/A N/A 0.88 Telonis [41] N/A 0.86 N/A Podolsky [44] 0.92 N/A N/A Cai [19] N/A 0.85 0.86 MLW-gcForest 0.92 0.87 0.86 LIHC Guo [20] N/A N/A 0.82 Telonis [41] N/A 0.90 N/A Tan [45] 0.77 0.83 N/A Friemel [46] N/A 0.87 N/A MLW-gcForest 0.91 0.87 0.85 GBM Guo [20] N/A N/A 0.78 Lu [21] 0.92 0.88 N/A Ryu [47] 0.83 0.80 N/A MLW-gcForest 0.87 0.89 0.86 STAD Liao [22] N/A 0.84 N/A Telonis [41] N/A 0.85 N/A MLW-gcForest 0.88 0.87 0.87…”

Section: Cancermentioning

confidence: 99%

“…Guo et al [20] proposed a hierarchical deep learning model to learn high-level representations in transcriptome data and gene expression data to classify cancer subtypes. Lu et al [21] developed a three-level machine-learning model to identify glioma subtypes. Liao et al [22] used a random forest method based on isomiR data to classify six cancers.…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

MLW-gcForest: A Multi-Weighted gcForest Model for Cancer Subtype Classification by Methylation Data

et al. 2019

View full text Add to dashboard Cite

Effective cancer treatment requires a clear subtype. Due to the small sample size, high dimensionality, and class imbalances of cancer gene data, classifying cancer subtypes by traditional machine learning methods remains challenging. The gcForest algorithm is a combination of machine learning methods and a deep neural network and has been indicated to achieve better classification of small samples of data. However, the gcForest algorithm still faces many challenges when this method is applied to the classification of cancer subtypes. In this paper, we propose an improved gcForest algorithm (MLW-gcForest) to study the applicability of this method to the small sample sizes, high dimensionality, and class imbalances of genetic data. The main contributions of this algorithm are as follows: (1) Different weights are assigned to different random forests according to the classification ability of the forests. (2) We propose a sorting optimization algorithm that assigns different weights to the feature vectors generated under different sliding windows. The MLW-gcForest model is trained on the methylation data of five data sets from the cancer genome atlas (TCGA). The experimental results show that the MLW-gcForest algorithm achieves high accuracy and area under curve (AUC) values for the classification of cancer subtypes compared with those of traditional machine learning methods and state of the art methods. The results also show that methylation data can be effectively used to diagnose cancer.

show abstract

Machine Learning–Based Radiomics for Molecular Subtyping of Gliomas

Cited by 237 publications

References 33 publications

Amino acid metabolism‐related gene expression‐based risk signature can better predict overall survival for glioma

Amino acid metabolism‐related gene expression‐based risk signature can better predict overall survival for glioma

A nuclear transport-related gene signature combined with IDH and 1p/19q better predicts the prognosis of glioma patients

MLW-gcForest: A Multi-Weighted gcForest Model for Cancer Subtype Classification by Methylation Data

Contact Info

Product

Resources

About