A Novel Computational Model to Predict Subcellular Protein Localizations
ISEF · 2019 Computational Biology and Bioinformatics
Overview
Approximately 20,000 different human proteins exist to perform specific tasks within a cell’s organelles focused on building, repairing, and giving signals needed for the body to survive. However, the majority of proteins have yet to be catalogued into a classification system due to lack of knowledge of their subcellular localizations. In 2018, consumer medications targeted only 3% of all proteins. These targeted proteins can occur in multiple areas of the body, damaging non-targeted regions in the form of side effects. This project focused on the development of a deep learning algorithm capable of classification of all possible subcellular localizations of unknown proteins based on fluorescent microscope image data. The classification model used a 34-layers-deep residual neural network for transfer learning. Many training parameters were optimized to deal with the image input, and the model performance was validated using 10-fold cross-validation. The resulting framework takes RGBY image input and produces a CSV file with predicted labels for the aforementioned images. The model was tested on 120,000 additional images and reached an F1 score of 79%, resulting in a 10% improvement over the previous state-of-the-art classifier developed by the Human Protein Atlas. Its ease-of-use and robustness are also improved over other technologies such as CellProfiler. This algorithm can facilitate the mapping of the human proteome, which could lead to developing more target-specific medications, nullifying harmful side effects caused by mislocalization of proteins, and allowing for earlier diagnoses of disease such as certain types of cancers and genetic disorders.
Competition history
- ISEF 2019
Resources
Related projects
ISEF · 2023
A Novel Phosphosite Localization Approach Using Tandem Phosphoprotein Mass Spectra and Temporal Dilated Networks
ISEF · 2018
Developing a Machine Learning Model to Identify Protein-Protein Interaction Hotspots to Facilitate Drug Discovery
ISEF · 2024
Utilizing Deep Learning to Facilitate Diagnosis of Look-Alike Leukemia Subtypes
ISEF · 2021
Application of Deep Learning in Target Identification Through Determining the Mechanism of Action Given Cellular Signature Data
ISEF · 2023
The Annotation of Novel Datasets for the Training of Cellori (Cell Origin) Spots, a Deep Learning Algorithm for RNA FISH Spot Detection
ISEF · 2024
AP Using Machine Learning Algorithms To Predict Whether Phytochemicals Are Cancerous or Anti-Cancerous
ISEF · 2015
Improved Cancer Detection and Diagnosis through a Novel Combination of Cell Segmentation and Artificial Intelligence Techniques
ISEF · 2023
A Novel Exosome-Encapsulated-miRNA-Based Machine Learning Model for Blood-Based Diagnosis, Subtype Classification, and Stage Detection of Multiple Cancer Types
Closest projects by meaning, across every fair and year in the corpus.
Browse more like this
Source: Regeneron International Science and Engineering Fair