DNA methylation-based classification of human central nervous system tumors [validation set]
Ontology highlight
ABSTRACT:
ORGANISM(S): Homo sapiens
PROVIDER: GSE109379 | GEO | 2018/01/18
REPOSITORIES: GEO
Similar Datasets
Project description:Background: Modern neuropathology is challenged by an increasing number of clinically-relevant CNS tumor subgroups that require assessment of a multitude of molecular markers for classification, as well a highly trained medical staff. Failure to meet this challenge leads to tumor misclassification, which can have severe consequences for affected patients. Methods: We compiled a cohort of genome-wide DNA methylation profiles of 2,682 tumors from 82 histologically and/or molecularly distinct CNS tumor classes across all ages and histologies that served as reference for a Random Forest-based diagnostic classifier. This classifier was used to prospectively investigate a further 1,104 CNS tumor samples in order to determine its clinical utility. Results: The classifier was able to reliably assign tumor samples to a given diagnostic category with a misclassification rate of less than 2%. The system functioned robustly across laboratories and using different DNA methylation profiling techniques. Prospective application to clinical samples resulted in a reclassification of 12% of tumors compared with standard practice alone. A further 12% could not be classified by methylation profiling – this subset was highly enriched for unusual syndrome-associated tumors and likely novel entities. Conclusion: This study represents a proof-of-concept for the application of machine learning approaches in molecular diagnostics using a single, easy-to-use assay. The reference cohort and Random Forest-based classifier are available online as a valuable community tool for improving precision in brain tumor diagnostics. We expect that approaches similar to the one presented herein will rapidly restructure diagnostic practice in neurooncology and across tumor pathology.
2018-01-14 | GSE90496 | GEO
Project description:The paper "Metabolomic Machine Learning Predictor for Diagnosis and Prognosis of Gastric Cancer" addresses the need for non-invasive diagnostic tools for gastric cancer (GC). Traditional methods like endoscopy are invasive and expensive. The authors conducted a targeted metabolomics analysis of 702 plasma samples to develop machine learning models for GC diagnosis and prognosis. The diagnostic model, using 10 metabolites, achieved a sensitivity of 0.905, outperforming conventional protein marker-based methods. The prognostic model effectively stratified patients into risk groups, surpassing traditional clinical models.
I have successfully reproduced the diagnosis model from the paper. This machine learning-based system differentiates GC patients from non-GC controls using metabolomics data from plasma samples analyzed by liquid chromatography-mass spectrometry (LC-MS). The model focuses on 10 metabolites, including succinate, uridine, lactate, and serotonin. Employing LASSO regression and a random forest classifier, the model achieved an AUROC of 0.967, with a sensitivity of 0.854 and specificity of 0.926. This model significantly outperforms traditional diagnostic methods and underscores the potential of integrating machine learning with metabolomics for early GC detection and treatment.
2024-07-23 | MODEL2407230001 | BioModels
Project description:Background: The development of a new drug from candidate to market is a complex process requiring vast resources of time, money and personnel. The rate of failure in the development pipeline is enormous, leading to wasted resources that could have been better employed on alternative candidates. The requirement for early stage prediction of toxicity is, then, of paramount importance to expedite the introduction of new therapies to clinical practice. To date, most transcriptomics efforts to solve this problem have applied Support Vector Machine techniques to data derived from in vivo studies in rats. Results: We applied a toxicogenomics approach to determine whether known renal toxicants could be identified as such from their effects on the transcriptome of the human renal proximal tubular epithelial cell line, HK-2. Based on clustering of differentially expressed genes, we identified 3 toxicity groups within the set of compounds. We used Random Forest to generate a classifier to accurately place compounds in groups. The classifier is based on a signature biomarker comprising 21 genes identified by Random Forest and could differentiate between the groups with high accuracy. Furthermore, we could correctly classify external samples from a dataset exhibiting a marked ‘batch effect’. Conclusions: No toxicity-associated gene expression alterations could be identified across a set of toxic compounds. Random Forest is a suitable technique for the classification of compounds into toxicity groups. Using a measure of differential expression rather than expression level per se generates a robust classifier that can potentially be applied in a platform-independent manner.
2015-12-31 | GSE27196 | GEO
Project description:Background: The development of a new drug from candidate to market is a complex process requiring vast resources of time, money and personnel. The rate of failure in the development pipeline is enormous, leading to wasted resources that could have been better employed on alternative candidates. The requirement for early stage prediction of toxicity is, then, of paramount importance to expedite the introduction of new therapies to clinical practice. To date, most transcriptomics efforts to solve this problem have applied Support Vector Machine techniques to data derived from in vivo studies in rats. Results: We applied a toxicogenomics approach to determine whether known renal toxicants could be identified as such from their effects on the transcriptome of the human renal proximal tubular epithelial cell line, HK-2. Based on clustering of differentially expressed genes, we identified 3 toxicity groups within the set of compounds. We used Random Forest to generate a classifier to accurately place compounds in groups. The classifier is based on a signature biomarker comprising 21 genes identified by Random Forest and could differentiate between the groups with high accuracy. Furthermore, we could correctly classify external samples from a dataset exhibiting a marked ‘batch effect’. Conclusions: No toxicity-associated gene expression alterations could be identified across a set of toxic compounds. Random Forest is a suitable technique for the classification of compounds into toxicity groups. Using a measure of differential expression rather than expression level per se generates a robust classifier that can potentially be applied in a platform-independent manner.
2015-12-31 | GSE27167 | GEO
Project description:Background: The development of a new drug from candidate to market is a complex process requiring vast resources of time, money and personnel. The rate of failure in the development pipeline is enormous, leading to wasted resources that could have been better employed on alternative candidates. The requirement for early stage prediction of toxicity is, then, of paramount importance to expedite the introduction of new therapies to clinical practice. To date, most transcriptomics efforts to solve this problem have applied Support Vector Machine techniques to data derived from in vivo studies in rats. Results: We applied a toxicogenomics approach to determine whether known renal toxicants could be identified as such from their effects on the transcriptome of the human renal proximal tubular epithelial cell line, HK-2. Based on clustering of differentially expressed genes, we identified 3 toxicity groups within the set of compounds. We used Random Forest to generate a classifier to accurately place compounds in groups. The classifier is based on a signature biomarker comprising 21 genes identified by Random Forest and could differentiate between the groups with high accuracy. Furthermore, we could correctly classify external samples from a dataset exhibiting a marked ‘batch effect’. Conclusions: No toxicity-associated gene expression alterations could be identified across a set of toxic compounds. Random Forest is a suitable technique for the classification of compounds into toxicity groups. Using a measure of differential expression rather than expression level per se generates a robust classifier that can potentially be applied in a platform-independent manner.
2015-12-31 | GSE27168 | GEO
Project description:Background: The development of a new drug from candidate to market is a complex process requiring vast resources of time, money and personnel. The rate of failure in the development pipeline is enormous, leading to wasted resources that could have been better employed on alternative candidates. The requirement for early stage prediction of toxicity is, then, of paramount importance to expedite the introduction of new therapies to clinical practice. To date, most transcriptomics efforts to solve this problem have applied Support Vector Machine techniques to data derived from in vivo studies in rats. Results: We applied a toxicogenomics approach to determine whether known renal toxicants could be identified as such from their effects on the transcriptome of the human renal proximal tubular epithelial cell line, HK-2. Based on clustering of differentially expressed genes, we identified 3 toxicity groups within the set of compounds. We used Random Forest to generate a classifier to accurately place compounds in groups. The classifier is based on a signature biomarker comprising 21 genes identified by Random Forest and could differentiate between the groups with high accuracy. Furthermore, we could correctly classify external samples from a dataset exhibiting a marked ‘batch effect’. Conclusions: No toxicity-associated gene expression alterations could be identified across a set of toxic compounds. Random Forest is a suitable technique for the classification of compounds into toxicity groups. Using a measure of differential expression rather than expression level per se generates a robust classifier that can potentially be applied in a platform-independent manner.
2015-12-31 | GSE27210 | GEO
Project description:Background: The development of a new drug from candidate to market is a complex process requiring vast resources of time, money and personnel. The rate of failure in the development pipeline is enormous, leading to wasted resources that could have been better employed on alternative candidates. The requirement for early stage prediction of toxicity is, then, of paramount importance to expedite the introduction of new therapies to clinical practice. To date, most transcriptomics efforts to solve this problem have applied Support Vector Machine techniques to data derived from in vivo studies in rats. Results: We applied a toxicogenomics approach to determine whether known renal toxicants could be identified as such from their effects on the transcriptome of the human renal proximal tubular epithelial cell line, HK-2. Based on clustering of differentially expressed genes, we identified 3 toxicity groups within the set of compounds. We used Random Forest to generate a classifier to accurately place compounds in groups. The classifier is based on a signature biomarker comprising 21 genes identified by Random Forest and could differentiate between the groups with high accuracy. Furthermore, we could correctly classify external samples from a dataset exhibiting a marked ‘batch effect’. Conclusions: No toxicity-associated gene expression alterations could be identified across a set of toxic compounds. Random Forest is a suitable technique for the classification of compounds into toxicity groups. Using a measure of differential expression rather than expression level per se generates a robust classifier that can potentially be applied in a platform-independent manner.
2015-12-31 | GSE27208 | GEO
Project description:Background: The development of a new drug from candidate to market is a complex process requiring vast resources of time, money and personnel. The rate of failure in the development pipeline is enormous, leading to wasted resources that could have been better employed on alternative candidates. The requirement for early stage prediction of toxicity is, then, of paramount importance to expedite the introduction of new therapies to clinical practice. To date, most transcriptomics efforts to solve this problem have applied Support Vector Machine techniques to data derived from in vivo studies in rats. Results: We applied a toxicogenomics approach to determine whether known renal toxicants could be identified as such from their effects on the transcriptome of the human renal proximal tubular epithelial cell line, HK-2. Based on clustering of differentially expressed genes, we identified 3 toxicity groups within the set of compounds. We used Random Forest to generate a classifier to accurately place compounds in groups. The classifier is based on a signature biomarker comprising 21 genes identified by Random Forest and could differentiate between the groups with high accuracy. Furthermore, we could correctly classify external samples from a dataset exhibiting a marked ‘batch effect'. Conclusions: No toxicity-associated gene expression alterations could be identified across a set of toxic compounds. Random Forest is a suitable technique for the classification of compounds into toxicity groups. Using a measure of differential expression rather than expression level per se generates a robust classifier that can potentially be applied in a platform-independent manner.
2015-12-31 | GSE27182 | GEO
Project description:Background: The development of a new drug from candidate to market is a complex process requiring vast resources of time, money and personnel. The rate of failure in the development pipeline is enormous, leading to wasted resources that could have been better employed on alternative candidates. The requirement for early stage prediction of toxicity is, then, of paramount importance to expedite the introduction of new therapies to clinical practice. To date, most transcriptomics efforts to solve this problem have applied Support Vector Machine techniques to data derived from in vivo studies in rats. Results: We applied a toxicogenomics approach to determine whether known renal toxicants could be identified as such from their effects on the transcriptome of the human renal proximal tubular epithelial cell line, HK-2. Based on clustering of differentially expressed genes, we identified 3 toxicity groups within the set of compounds. We used Random Forest to generate a classifier to accurately place compounds in groups. The classifier is based on a signature biomarker comprising 21 genes identified by Random Forest and could differentiate between the groups with high accuracy. Furthermore, we could correctly classify external samples from a dataset exhibiting a marked ‘batch effect’. Conclusions: No toxicity-associated gene expression alterations could be identified across a set of toxic compounds. Random Forest is a suitable technique for the classification of compounds into toxicity groups. Using a measure of differential expression rather than expression level per se generates a robust classifier that can potentially be applied in a platform-independent manner.
2015-12-31 | GSE27170 | GEO
Project description:Background: The development of a new drug from candidate to market is a complex process requiring vast resources of time, money and personnel. The rate of failure in the development pipeline is enormous, leading to wasted resources that could have been better employed on alternative candidates. The requirement for early stage prediction of toxicity is, then, of paramount importance to expedite the introduction of new therapies to clinical practice. To date, most transcriptomics efforts to solve this problem have applied Support Vector Machine techniques to data derived from in vivo studies in rats. Results: We applied a toxicogenomics approach to determine whether known renal toxicants could be identified as such from their effects on the transcriptome of the human renal proximal tubular epithelial cell line, HK-2. Based on clustering of differentially expressed genes, we identified 3 toxicity groups within the set of compounds. We used Random Forest to generate a classifier to accurately place compounds in groups. The classifier is based on a signature biomarker comprising 21 genes identified by Random Forest and could differentiate between the groups with high accuracy. Furthermore, we could correctly classify external samples from a dataset exhibiting a marked ‘batch effect’. Conclusions: No toxicity-associated gene expression alterations could be identified across a set of toxic compounds. Random Forest is a suitable technique for the classification of compounds into toxicity groups. Using a measure of differential expression rather than expression level per se generates a robust classifier that can potentially be applied in a platform-independent manner.
2015-12-31 | GSE27169 | GEO