Honors Program Theses

Award/Availability

Open Access Honors Program Thesis

First Advisor

Aleksandar Poleksic

Abstract

Protein function prediction is a crucial task in bioinformatics, traditionally approached through sequence-based, structure-based, or interaction network models. These models, however, fail to capture transient conformations that occur during large-scale conformational changes. MorphToGO proposes a novel methodology of function prediction with the development of a video classification model to analyze morph animations. The image classification portion of the model utilizes the pretrained EfficientNetV2 model to extract features from animation frames, and the video portion consists of a modified GRU architecture for sequence-to-sequence embedding regression to predict the protein’s functional semantics. A custom bidirectional cosine loss function was developed to handle unordered, variable-length output sets. Model evaluation demonstrated high semantic similarity between predicted and ground truth embeddings, which suggests high potential for identifying functions from transitional states.

Year of Submission

2025

Department

Department of Computer Science

University Honors Designation

A thesis submitted in partial fulfillment of the requirements for the designation University Honors

Date Original

2025

Object Description

1 PDF file (20 pages)

Creative Commons License

Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International License
This work is licensed under a Creative Commons Attribution-NonCommercial-Share Alike 4.0 International License.

Language

en

File Format

application/pdf

Download

Comments?

Share

COinS