Code Repository

Application of Machine Learning Methods to Predict Phytoplankton Blooms and Determine Microbial Biomarkers Using Marine Microbiomes

Code developed during the TFM entitled Application of Machine Learning Methods to Predict Phytoplankton Blooms and Determine Microbial Biomarkers Using Marine Microbiomes can be found in the code folder that contains:

Original_files_preprocess:

• Data dowload from SRA:
  • Data_collection.md
  • sra_explorer_download.sh
• Environmental metadata and sequence files preparation:
  • Metadata_files_preprocess.md
• Raw sequences preprocess:
  • Sequences_preprocess.md
• ASVs calculation:
  • ASVs_calculation.md

Chlorophyll_groups:

• PCAs of environmental variables:
  • PCA_environmental_data.R
• Data distribution according to different percentiles of total chlorophyll concentrations:
  • Percentiles.R

Main folder scripts:

• Data preparation:

  • Preparation of DADA2 results: DADA2_tables.R
  • Merge of Envision and Dimension environmental metadata and sample labelling: Initial_preprocess.R

• Data exploration:

  • ASVs without filtering exploration: ASVs_no_filter_exploration.R
  • ASVs filtering and exploration: ASVs_filter_exploration.R
  • Clusters generation and exploratory analyses: Clusters_grouping_exploration.R

• Train, validation with 5 and 10 k-fold repeated cross-validation and performance test of Random Forest and SVM models:

  • Random Forest with Clusters: Clusters_model_1.R
  • Random Forest with Filtered ASVs: ASVs_model_1.R
  • Random Forest with Unfiltered ASVs: ASVs_no_filter_model_1.R
  • SVM Radial with Filtered ASVs and validation results: ASVs_SVM_model.r
  • Validation and performance metrics for RF and SVM models: RF_SVM_models_results.R

• Synthetic data:

  • Data generation and models training: ASVs_synthetic_RF.R
  • Performance results of RF model with synthetic data: ASVs_synthetic_RF_results.R

• Features importance: feature_importance.R