Genes (Dec 2021)

Machine Learning and Bioinformatics Framework Integration to Potential Familial DCM-Related Markers Discovery

  • Concetta Schiano,
  • Monica Franzese,
  • Filippo Geraci,
  • Mario Zanfardino,
  • Ciro Maiello,
  • Vittorio Palmieri,
  • Andrea Soricelli,
  • Vincenzo Grimaldi,
  • Enrico Coscioni,
  • Marco Salvatore,
  • Claudio Napoli

DOI
https://doi.org/10.3390/genes12121946
Journal volume & issue
Vol. 12, no. 12
p. 1946

Abstract

Read online

Objectives: Dilated cardiomyopathy (DCM) is characterized by a specific transcriptome. Since the DCM molecular network is largely unknown, the aim was to identify specific disease-related molecular targets combining an original machine learning (ML) approach with protein-protein interaction network. Methods: The transcriptomic profiles of human myocardial tissues were investigated integrating an original computational approach, based on the Custom Decision Tree algorithm, in a differential expression bioinformatic framework. Validation was performed by quantitative real-time PCR. Results: Our preliminary study, using samples from transplanted tissues, allowed the discovery of specific DCM-related genes, including MYH6, NPPA, MT-RNR1 and NEAT1, already known to be involved in cardiomyopathies Interestingly, a combination of these expression profiles with clinical characteristics showed a significant association between NEAT1 and left ventricular end-diastolic diameter (LVEDD) (Rho = 0.73, p = 0.05), according to severity classification (NYHA-class III). Conclusions: The use of the ML approach was useful to discover preliminary specific genes that could lead to a rapid selection of molecular targets correlated with DCM clinical parameters. For the first time, NEAT1 under-expression was significantly associated with LVEDD in the human heart.

Keywords