Evaluation of miRNA-based classifiers for cancer diagnosis
Cancers account for the major deadliest noncommunicable diseases across all segments of the population and responsible for around 13% of all deaths world-wide. Cancer prevalence rate has noticeably quickened its pace in Malaysia and the world as we know it. Conventional diagnostic imaging and invasi...
| Main Authors: | , , |
|---|---|
| Format: | Conference or Workshop Item |
| Language: | English English |
| Published: |
2017
|
| Subjects: | |
| Online Access: | http://irep.iium.edu.my/56372/ http://irep.iium.edu.my/56372/ http://irep.iium.edu.my/56372/19/56372-abstract%20book.pdf http://irep.iium.edu.my/56372/3/ICETAS.pdf |
| Summary: | Cancers account for the major deadliest noncommunicable diseases across all segments of the population and responsible for around 13% of all deaths world-wide. Cancer prevalence rate has noticeably quickened its pace in Malaysia and the world as we know it. Conventional diagnostic imaging and invasive biopsy examinations are still the gold standard for the diagnosis of cancer. However, these conventional methods suffer from low diagnosis sensitivity compounded by work-intensive analysis. There have indeed been a number of miRNA studies to tackle the challenges associated with cancer biomarker discovery. However, the existing diagnosis techniques using miRNA suffer from low diagnosis accuracy, sensitivity, and specificity. The low diagnosis accuracy and sensitivity of the existing techniques stems from the fact that there is extremely low miRNA count in body fluids and the presence of a huge number of irrelevant miRNAs in the expression data. There is also an inevitable problem of cross contamination between cells and exosomes in sample preparation steps. This paper describes the state-of-the-art miRNA-based classifiers for cancer miRNA expression classification. To lower the computational complexity, we employ a heuristic-based miRNA selection approach to select relevant miRNAs that are directly responsible for cancer diagnosis. Among the classifiers, Random Forest (RF) has achieved an average accuracy of 97% over 11 independent datasets. The experimental results are quite encouraging and the predictive framework managed to classify cancer accurately even with much noise contaminated in the datasets. |
|---|