Research

PhD Dissertation

Developed and modified different ensemble based machine learning and deep learning algorithms where prediction accuracy of anti-cancer drug sensitivity got improved compared to other traditional methods.
Collected and analyzed big pharmacogenomics (genomic, trascriptomic, proteomic and so on) databases to detect the inconsistency of data generation, and then implemented transfer learning approaches to reduce computational complexity of modeling along with improving predictive accuracy and data augmentation.
Developed a novel approach for Multi-Dimensional-Scale (MDS) image generation using drug target properties and used these images in Convolutional Neural Network (CNN) for better recommender system in drug employment.
Created a mathematical framework for analyzing drug combination toxicity for personalized medicine applications.
Enhanced a Sequential Feature Selection and Inference approach using Multivariate Random Forests by considering linear relationship among target features.
Designed Probabilistic Random Forest with Bayesian Networks, Expectation Maximization and Mixture Modeling for Anticancer Drug Sensitivity Prediction.
Developed a dose-response curve predictive model using functional characteristics of genomic characterization and pharmacogenomics data.
Considering heterogeneity of different cancer subtypes, developed drug response predictive model which can also classify drug-sensitive and -insensitive cell lines.

Created a framework where different genomic characterization datasets can be integrated for Drug Sensitivity Prediction Model Refinement for both univariate and multivariate cases.
Created three R-packages (widely used) for improving predictive accuracy of large scale databases by integrated modeling.

Statistical analysis of mouse genome sequence using Trinity.
Constructed stochastic models of allelopathic interactions between two competing phytoplankton species as a continuous time Markov chain model.
Using Simplex algorithm and Linear Programming along with GUROBI and AMPL, constructed a production cost optimization model, which reduced cost by 5% compared to other models.
Live Handwritten Character Recognition using Convolutional Neural Network (CNN).

Developed a novel algorithm for direct strain estimation of Ultrasound Elastography.
Completed other projects such as ’Designing Parity Generator or Checker using Cadence Virtuoso Software’, ’Line Follower Robot’, ’Stepper Motor control from PC via USB’, ’8 bit Digital Computer design and simulation’ and more.