Legrand, Damien (2024) Generative models for adaptative immunity PRE - Research Project, ENSTA.
![]() | PDF Restricted to Registered users only 621Kb |
Abstract
This raport investigates the generation of protein sequences using Long Short-Term Memory (LSTM) networks and Variational Autoencoders (VAE). The study compares these models based on their ability to produce biologically meaningful and diverse protein sequences. The goal is to find relevant metrics to assess these generative models. To do so, I am going to evaluate a range of metrics over real data before using them to assess LSTM and VAE generations. Unfortunately, it was not possible to generate sequences with these models, thus I evaluate the metrics over mutated sequences to determine at what point the sequences lose the grammar of immune sequences due to mutations, which act as a proxy.
Item Type: | Thesis (PRE - Research Project) |
---|---|
Uncontrolled Keywords: | Protein sequences, Generative models, Synthetic data, Metrics |
Subjects: | Information and Communication Sciences and Technologies |
ID Code: | 10119 |
Deposited By: | Damien LEGRAND |
Deposited On: | 28 août 2024 18:45 |
Dernière modification: | 28 août 2024 18:45 |
Repository Staff Only: item control page