SIB

Authors:

• António Duarte | PG45464
• Roberto Bullitta | PG45474
• Vânia Miguel | PG45971

Repository Guide:

• Files (folder) -> contains files generated during the development of this project (.txt and .csv); the following files could not be uploaded due to their large size:
  • data_train.csv -> training data (after feature extraction)
  • data_test.csv -> test data (after feature extraction)
  • X_train_sc.csv -> scaled training data (minmax scaler)
  • X_train_sc_z.csv -> scaled training data (standard scaler)
• Experimental (folder) -> contains experimental procedures (.ipynb)
  • GBR_clust.ipynb (file) -> train distinct models based on a previous clustering analysis
  • MLP_hyper.ipynb (file) -> optimize MLP hyper-parameters
• feature_extraction.py (file) -> contains functions for feature extraction
• C_P_FE.ipynb (file) -> context, preprocessing and feature extraction
• AE.ipynb (file) -> exploratory analysis
• UL.ipynb (file) -> unsupervised learning
• SL.ipynb (file) -> supervised learning
• DL_MLP.ipynb (file) -> deep learning (multilayer perceptron)
• DL_CNN.ipynb (file) -> deep learning (convolutional neural network)

Content Guide:

• Contextualization
• Preprocessing
• Feature Extraction
• Exploratory Analysis
• Unsupervised Learning
• Supervised Learning (shallow learning)
• Deep learning

Contextualization:

The data used in this report was retrived from KAGGLE: https://www.kaggle.com/c/novozymes-enzyme-stability-prediction
Each data example consists of a protein sequence, a pH value and thermostability index. Predicting the thermostability is fundamental in enzyme engeneering for a wide variety of applications. Employing ML techniques is of great value to achieve the latter purpose as it saves time and money.

Preprocessing:

• Data processing
• Identification of the train and test datasets
• Treatment of Missing values

Feature Extraction

• Identification of the Descriptors with BioPython and ProPy

Exploratory Analysis

• Outlier Treatment (dependent variable)
• Standardization
• Thermostability value distribution
• Correlation Analysis (Pearson, Spearman, f_regression, mutual_information)
• Multicollinearity Analysis

Unsupervised Learning

• PCA
• tSNE
• K-means

Supervised Learning

• Cross-validation, Hyper-parameter tuning, and Model selection
• Linear Regression
• K-Nearest Neighbors
• Support Vector Machine
• Random Forest
• Adaptive Boosting
• Gradient Boosting

Deep learning

• MLP
• CNN

Credits:

• Curricular Unit Slides
• Deep Learning: https://machinelearningmastery.com/grid-search-hyperparameters-deep-learning-models-python-keras/