Loading...

Enhancing Compound and Gene Image-based Profiling for Drug Discovery and Validation based on Structural/Computational Methods

Talaei, Tahereh | 2022

174 Viewed
  1. Type of Document: M.Sc. Thesis
  2. Language: Farsi
  3. Document No: 55354 (19)
  4. University: Sharif University of Technology
  5. Department: Computer Engineering
  6. Advisor(s): Rohban, Mohammad Hossein; Kalhor, Hamid Reza
  7. Abstract:
  8. The image-based profile is a technology by which image morphology information is transformed into a multidimensional profile from a set of image-derived features. These profiles can be used to extract biologically meaningful biological information. For example, in the drug discovery process, the mechanism of action of a drug or disease can be identified by examining the morphological properties of the drug in the patient’s cell or tissue and used to design new drugs or use existing drugs for various diseases. High-throughput imaging technology allows the imaging of a large number of different experiments. Extracting valuable features and a good representation of features is the main challenge of this issue. There are traditional methods and machine learning methods that have been studied for feature extraction. Given that the existing data do not have definitive labels, one of the main challenges in this issue has always been the lack of labeled data. In this project, we have used self-supervised learning to learn the display of cells. The CPJUMP1 dataset we used in this project has been introduced recently, allowing for the first time comparison between chemical compounds and genetic perturbation. Therefore, in this project, we examined the similarities of different drugs and genes to evaluate the results more accurately. Then, the relationships between chemical compounds and target proteins have been investigated using molecular docking software. Our results in learning representations for these data compared to existing methods Have improved results
  9. Keywords:
  10. Cellular Morphology ; Imaged Based Profiling ; Molecular Docking ; Self-Supervised Learning

 Digital Object List

 Bookmark