antibodyfvtm50predictor

MRes project - Predicting the melting temperature of antibodies using machine learning techniques.

https://github.com/michaelskehan/antibodyfvtm50predictor

Science Score: 44.0%

This score indicates how likely this project is to be science-related based on various indicators:

  • CITATION.cff file
    Found CITATION.cff file
  • codemeta.json file
    Found codemeta.json file
  • .zenodo.json file
    Found .zenodo.json file
  • DOI references
  • Academic publication links
  • Academic email domains
  • Institutional organization owner
  • JOSS paper metadata
  • Scientific vocabulary similarity
    Low similarity (6.0%) to scientific vocabulary
Last synced: 6 months ago · JSON representation ·

Repository

MRes project - Predicting the melting temperature of antibodies using machine learning techniques.

Basic Info
  • Host: GitHub
  • Owner: MichaelSkehan
  • Language: Jupyter Notebook
  • Default Branch: main
  • Size: 28.2 MB
Statistics
  • Stars: 0
  • Watchers: 1
  • Forks: 0
  • Open Issues: 0
  • Releases: 0
Created about 2 years ago · Last pushed about 2 years ago
Metadata Files
Readme Citation

README.md

AntibodyFvTm50Predictor

MRes project - prediction the melting temperature of antibodies using machine learning techniques

In this work, AntiBERTy was used to encode antibody sequences. A 2D plot generated using t-SNE shows distinct separation between species based on the 512-feature encoding and a reduced, 60-feature encoding. Using a Gradient Boosting Tree (GBT) method on a blind test dataset, the model showed a Pearson Correlation Coefficient of 0.83 and mean absolute error of 2.69°C between predicted and actual Tm50.

The results from this study demonstrate that, firstly, AntiBERTy can capture unique, species specific features from amino acid sequences, and secondly, an accurate Tm50 predictor can be achieved using a GBT model trained on AntiBERTy encoded amino acid sequences.

A regression model using GBT was shown to be effective in predicting Tm50 with a MAE of 2.61C, PCC 0.84. GBT_act_vs_pred_2

AntiBERTy can capture unique, species specific features from amino acid sequences, even after feature reduction form 512-features to 60.

512-features

TSNE2D_AbYsis_full_berty_per50

60-features

TSNE2D_AbYsis_60_berty_per50

Currently, the webapp is not hosted online but can be trialled using Flask to run on a local server using flask --app main run when in the webapp directory. webapp

Owner

  • Login: MichaelSkehan
  • Kind: user

Citation (CITATION.cff)

cff-version: 1.2.0
message: "Citation:"
authors:
  - family-names: Skehan
    given-names: Michael
title: "AntibodyFvTm50Predictor"
version: 1.0.0
date-released: 2023-08-29

GitHub Events

Total
Last Year

Dependencies

requirements.txt pypi
  • Flask ==2.2.2
  • Flask_Mail ==0.9.1
  • biovec ==0.2.7
  • igfold ==0.3.1
  • ipython ==8.6.0
  • joblib ==1.2.0
  • keras ==2.13.1
  • matplotlib ==3.6.2
  • numpy ==1.23.4
  • pandas ==1.5.1
  • python-dotenv ==1.0.0
  • scikit_learn ==1.2.2
  • scipy ==1.11.2
  • seaborn ==0.12.2
  • tensorflow ==2.13.0
  • torch ==2.0.0