پديدآورندگان :
Sadeghi H Department of Genetics, University of Science and Culture, ACECR; Reproductive Biomedicine Research Center, Royan Institute for
Reproductive Biomedicine, ACECR, Tehran, Iran , Sharifi Zarchi A 3Department of Stem Cells and Developmental Biology, Cell Science Research Center,
Royan Institute for Stem Cell Biology and Technology, ACECR, Tehran, Iran , Kamal A Department of Mathematical Sciences, Sharif University of Technology, Tehran, Iran , Shayesteh Pour B Department of Genetics, University of Science and Culture, ACECR; Reproductive Biomedicine Research Center, Royan Institute for
Reproductive Biomedicine, ACECR, Tehran, Iran , Gourabi H Department of Genetics, Reproductive Biomedicine Research Center, Royan Institute for
Reproductive Biomedicine, ACECR, Tehran, Iran , Keller A Department of Clinical Bioinformatics, Saarland University, Saarbr¨ ucken, Germany , Totonchi M m.totonchi@royaninstitute.org Department of Genetics, Reproductive Biomedicine Research Center, Royan Institute for
Reproductive Biomedicine, ACECR, Tehran, Iran
چكيده فارسي :
microRNAs (miRNAs) are a class of non-coding RNAs that regulate many cellular processes including tumorigenesis. Circulating miRNAs are known as less invasive markers in many malignancies such as cancer. Recent studies have shown that some specific miRNAs are deregulated in blood of early stage cancer patients compared to healthy controls. In this study, we aim to design subsets of circulating miRNAs can detect each type of cancer from unaffected controls and other types of cancers with high accuracy.
We used miRNA expression profiles from the cancer genome atlas (TCGA) and analyzed 6104 next-generation sequencing (NGS) data related to 14 different types of cancer tissues encompassing 5493 cancer samples and 611 healthy controls. We were using feature selection algorithm and support vector machine with 10 fold cross validation as machine learning method for improving detection accuracy.
By focusing on five miRNAs, we could separate all cancer samples from all normal samples with 97% accuracy. We obtained subsets with maximum 5 members and also acceptable accuracy for each cancer type. The highest accuracy received for thyroid carcinoma (98%) and kidney renal clear cell carcinoma (97%) with subset of three and two miRNAs, respectively. We also could classify samples in 3 classes (breast invasive carcinoma, normal breast tissue and all other normal and cancer tissues) just with 3 miRNAs.
Using these bioinformatics approach, we identified various subsets of miRNAs that could distinguish every type of cancer from unaffected controls. These subsets have potential to be evaluated in blood samples of each cancer type.
