Our Publications

Uncovering key molecular mechanisms in the early and late-stage of papillary thyroid carcinoma using association rule mining algorithm

Objective

Thyroid Cancer (TC) is the most frequent endocrine malignancy neoplasm. It is the sixth cause of cancer in women worldwide. The treatment process could be expedited by identifying the controlling molecular mechanisms at the early and late stages, which can contribute to the acceleration of treatment schemes and the improvement of patient survival outcomes. In this work, we study the significant mRNAs through Machine Learning Algorithms in both the early and late stages of Papillary Thyroid Cancer (PTC).

Method

During the course of our study, we investigated various methods and techniques to obtain suitable results. The sequence of procedures we followed included organizing data, using nested cross-validation, data cleaning, and normalization at the initial stage. Next, to apply feature selection, a t-test and binary Non-Dominated Sorting Genetic Algorithm II (NSGAII) were chosen to be employed. Later on, during the analysis stage, the discriminative power of the selected features was evaluated using machine learning and deep learning algorithms. Finally, we considered the selected features and utilized Association Rule Mining algorithm to identify the most important ones for improving the decoding of dominant molecular mechanisms in PTC through its early and late stages.

Decoding the Possible Molecular Mechanisms in Pediatric Wilms Tumor and Rhabdoid Tumor of the Kidney through Machine Learning Approaches

Objective: Wilms tumor (WT) and Rhabdoid tumor (RT) are pediatric renal tumors and their differentiation is based on histopathological and molecular analysis. The present study aimed to introduce the panels of mRNAs and microRNAs involved in the pathogenesis of these cancers using deep learning algorithms.

Methods: Filter, graph, and association rule mining algorithms were applied to the mRNAs/microRNAs data. Results: Candidate miRNAs and mRNAs with high accuracy (AUC: 97%/93% and 94%/97%, respectively) could differentiate the WT and RT classes in training and test data. Let-7a-2 and C19orf24 were identified in the WT, while miR-199b and RP1-3E10.2 were detected in the RT by analysis of Association Rule Mining.

Conclusion: The application of the machine learning methods could identify mRNA/miRNA patterns to discriminate WT from RT. The identified miRNAs/mRNAs panels could offer novel insights into the underlying molecular mechanisms that are responsible for the initiation and development of these cancers. They may provide further insight into the pathogenesis, prognosis, diagnosis, and molecular-targeted therapy in pediatric renal tumors.

Unraveling the link between PTBP1 and severe asthma through machine learning and association rule mining method

Severe asthma is a chronic inflammatory airway disease with great therapeutic challenges. Understanding the genetic and molecular mechanisms of severe asthma may help identify therapeutic strategies for this complex condition. RNA expression data were analyzed using a combination of artificial intelligence methods to identify novel genes related to severe asthma. Through the ANOVA feature selection approach, 100 candidate genes were selected among 54,715 mRNAs in blood samples of patients with severe asthmatic and healthy groups. A deep learning model was used to validate the significance of the candidate genes. The accuracy, F1-score, AUC-ROC, and precision of the 100 genes were 83%, 0.86, 0.89, and 0.9, respectively. To discover hidden associations among selected genes, association rule mining was applied. The top 20 genes including the PTBP1, RAB11FIP3, APH1A, and MYD88 were recognized as the most frequent items among severe asthma association rules. The PTBP1 was found to be the most frequent gene associated with severe asthma among those 20 genes. PTBP1 was the gene most frequently associated with severe asthma among candidate genes. Identification of master genes involved in the initiation and development of asthma can offer novel targets for its diagnosis, prognosis, and targeted-signaling therapy.

Key therapeutic targets implicated at the early stage of hepatocellular carcinoma identified through machine-learning approaches

Hepatocellular carcinoma (HCC) is the most frequent type of primary liver cancer. Early-stage detection plays an essential role in making treatment decisions and identifying dominant molecular mechanisms. We utilized machine learning algorithms to find significant mRNAs and microRNAs (miRNAs) at the early and late stages of HCC. First, pre-processing approaches, including organization, nested cross-validation, cleaning, and normalization were applied. Next, the t-test/ANOVA methods and binary particle swarm optimization were used as a filter and wrapper method in the feature selection step, respectively. Then, classifiers, based on machine learning and deep learning algorithms were utilized to evaluate the discrimination power of selected features (mRNAs and miRNAs) in the classification step. Finally, the association rule mining algorithm was applied to selected features for identifying key mRNAs and miRNAs that can help decode dominant molecular mechanisms in HCC stages. The applied methods could identify key genes associated with the early (e.g., Vitronectin, thrombin-activatable fibrinolysis inhibitor, lactate dehydrogenase D (LDHD), miR-590) and late-stage (e.g., SPRY domain containing 4, regucalcin, miR-3199-1, miR-194-2, miR-4999) of HCC. This research could establish a clear picture of putative candidate genes, which could be the main actors at the early and late stages of HCC.

A diagnostic miRNA panel to detect recurrence of ovarian cancer through artificial intelligence approaches

Background

Ovarian Cancer (OC) is the deadliest gynecology malignancy, whose high recurrence rate in OC patients is a challenging object. Therefore, having deep insights into the genetic and molecular mechanisms of OC recurrence can improve the target therapeutic procedures. This study aimed to discover crucial miRNAs for the detection of tumor recurrence in OC by artificial intelligence approaches.

Method

Through the ANOVA feature selection method, we selected 100 candidate miRNAs among 588 miRNAs. For their classification, a deep-learning model was employed to validate the significance of the candidate miRNAs. The accuracy, F1-score (high-risk), and AUC–ROC of classification test data based on the 100 miRNAs were 73%, 0.81, and 0.65, respectively. Association rule mining was used to discover hidden relations among the selected miRNAs.

Result

Five miRNAs, including miR-1914, miR-203, miR-135a-2, miR-149, and miR-9–1, were identified as the most frequent items among high-risk association rules. The identified miRNAs may target genes/proteins involved in epithelial–mesenchymal transition (EMT), resistance to therapy, and cancer stem cells; being responsible for the heterogeneity and plasticity of the tumor. Our conclusion presents mir-1914 as the significant candidat…