Prediction of tumor location in prostate cancer tissue using a machine learning system on gene expression data

Abstract

Background: Finding the tumor location in the prostate is an essential pathological step for prostate cancer diagnosis and treatment. The location of the tumor -the laterality -can be unilateral (the tumor is affecting one side of the prostate), or bilateral on both sides. Nevertheless, the tumor can be overestimated or underestimated by standard screening methods. In this work, a combination of efficient machine learning methods for feature selection and classification are proposed to analyze gene activity and select them as relevant biomarkers for different laterality samples. Results: A data set that consists of 450 samples was used in this study. The samples were divided into three laterality classes ( left, right, bilateral). The aim of this work is to understand the genomic activity in each class and find relevant genes as indicators for each class with nearly 99% accuracy. The system identified groups of differentially expressed genes (RTN1, HLA-DMB, MRI1) that are able to differentiate samples among the three classes. Conclusion: The proposed method was able to detect sets of genes that can identify different laterality classes. The resulting genes are found to be strongly correlated with disease progression. HLA-DMB and EIF4G2, which are detected in the set of genes can detect the left laterality, were reported earlier to be in the same pathway called Allograft rejection SuperPath.

Novel deep learning-based noise reduction technique for prostate magnetic resonance imaging

“…DL has been used for segmentation of the prostate and urethra [ 14 ], for detection of prostate cancer [ 15 ], and for radiation treatment planning for prostate cancer [ 16 ]. However, to our knowledge, there are no reports of the application of DL or artificial intelligence (AI) techniques to improve image quality and evaluate the anatomy and tumors in patients with prostate cancer.…”

Identification of novel hub genes and lncRNAs related to the prognosis and progression of pancreatic cancer by microarray and integrated bioinformatics analysis

“…It has also been shown to promote PC progression and gemcitabine resistance via a HOTTIP-HOXA13 axis route (44). RTN1 (reticulon 1) belongs to the reticulonencoding gene family and is considered as a specific marker and potential therapeutic target for non-small-cell lung cancers, colorectal cancer, and prostate cancer (45)(46)(47)(48). CHRAC1 (Chromatin Accessibility Complex Subunit 1) is a histone-fold protein-encoding gene, functioning in DNA transcription, replication, and packaging as a sequencespecific DNA binding component, and associated with survival progression among breast cancer patients (49,50).…”

Artificial Intelligence and Machine Learning in Prostate Cancer Patient Management—Current Trends and Future Perspectives

“…Liu et al [ 74 ] identified 12 CpG (cytosine and guanine on a genome) site markers and 13 promoter markers, using a deep neural network model, from an initial pool of 139,422 CpG sites, and the promoter methylation data contained 15,316 promoters and applied three machine learning strategies (moderated t-statistics, LASSO, and random-forest).This might further be used for liquid biopsy of cancers. Lately, Hamzeh et al [ 75 ] used a combination of efficient machine learning methods (Support Vector Machine (SVM)-Radial basis function kernel (SVM-RBF), Naive Bayes, Random Forest) to analyze gene activity and to identify the genes for the presence of PCa (on one side or both sides). The highest accuracy and precision for the different classifiers came from the SVM-RBF classifier, which was able to separate the different locations by an accuracy of 99% and found genes (HLA-DMB and EIF4G2) that are correlated with PCa progression.…”