Implementation of machine learning algorithms for medulloblastoma classification in a local galaxy server
Year: 2018
Type: other
Abstract: Medulloblastoma is the most common malignant brain tumor in children, comprising around 20% of all childhood brain tumors. It has been classified in four main molecular subgroups with different survival outcomes: WNT, SHH, group 3 and group 4. Currently, there is no gold-standard method for diagnostic of these molecular subgroups, which could be efficiently applied in clinical practice. Northcott and colleagues proposed a 22-gene panel for molecular classification of medulloblastoma from formalin-fixed paraffin embedded (FFPE) tissue blocks applying the NanoString nCounter® technology and the Prediction Analysis of Microarrays (PAM) algorithm. In their work, PAM was able to confidently predict around 88% of FFPE test samples using a training set of 101 and a validation set of 130 fresh-frozen samples. This pipeline has been applied at Barretos Cancer Hospital (BCH), Brazil, with a training set and additional genes have been tested in order to improve prognosis prediction. Moreover, Support Vector Machine (SVM) method has been developed for application in cases with inconclusive results for PAM algorithm. These R script pipelines were incorporated into galaxy tools in a local server using Planemo utilities. This implementation may automate and make medulloblastoma classification more accessible for clinicians and biologists in clinical practice, leading to tailored targeted therapies on clinical trials in near feature.
Institution: Hospital de Câncer de Barretos
Language: English
Subfield: Genetics (clinical)
Field: Medicine
Domain: Health Sciences
Open Access status: gold
OpenAlex ID: W2978067582