Integration of meta-analysis and supervised machine learning for pattern recognition in breast cancer using epigenetic data

Breast cancer is one of the most widespread diseases with high incidence and mortality rate in females. The accurate biomarker discovery for the early detection of patients prone to breast cancer is crucial in the treatment and diagnosis of breast cancer. The current study employed a comprehensive approach to detect an epigenomic data pattern of breast cancer using meta-analysis and machine learning approaches. Meta-analysis is a precise method that combines the results of multiple experiments. On the other hand, integrating and combining the test results through machine learning algorithms can deal with data complexity and heterogeneity. The main purpose of the current study was to discover the patterns of epigenome changes in the treatment and prognosis of breast cancer. NCBI and EBI databases were searched for ChIP-Seq data regarding the effect of the drugs on breast cancer. There were ten investigations carried out, four of which were appropriate meta-analysis. NOV, JUN and ZBTB7A transcription factors were identified as the biomarkers of breast cancer. Finally, pattern recognition was performed using nine different attribute weighting algorithms. Fourteen genes were selected by the majority of attribute weighting algorithms as the most informative genes including KIP, TCF12, ABCC5, HDAC11, IPP, HIST1H2AM, ZNF33B, PHF2, ELAVL3, TBC1D9B, TMEM217, CD34, ARHGEF26, and CENPL. The selected genes play vital roles in the occurrence of neoplasms and breast cancer. In this study, using a combination of meta-analysis and data mining, more comprehensive and reliable information were derived compared to the individual studies.