Pathway_gene_prediction_in_tomato

Scripts for our paper: Wang P, Moore BM, Ugyun S, Lehti-Shiu M, Barry C, Shiu SH. Optimizing the use of gene expression data to predict plant metabolic pathway memberships. New Phytologist

Expression datasets please see https://zenodo.org/record/4585635#.YEKEjMhvrrM

• File "Cufflink_matrix_sample_name_20180123.txt" contains the original FPKM for each sample replicate
• File "HTseq_matrix_sample_name_20180123.txt" contains the original read counts for each sample replicate
• File "Tomato_TPM.txt" contains the original TPM for each sample replicate
• Files starting with "Results_Fold_changes_", "Results_median_FPKM_" or "TPM_" contain FC, FPKM or TPM values for 41 datasets used in the manuscript.

Get expression matrix

Fold changes

• Rscript Get_fold_change_values.r

Median FPKM among replicates

• Rscript Get_median_FPKM_among_replicates.r

Median TPM among replicates

• Rscript Get_TPM.r

Calculate the gene-to-gene co-expression matrix

Partial correlation

• Rscript Corpcor.r expression_matrix

Note that, if the Coexpression_multiclass.py is used, you don't have to calculate the PCC, Spearman's and Mutual information for the gene-to-gene coexpression matrix using the following three scripts

Pearson correlation coefficient

• python PCC_pandas_20180918.py -file expression_matrix -path path_to_expression_matrix

Spearman's rank correlation coefficient

• python Spearman_pandas_20180918.py -file expression_matrix -path path_to_expression_matrix

Mutual information

• python MI_pandas_20180919.py -file expression_matrix -path path_to_expression_matrix -start where_the_subset_starts -stop where_the_subset_stops

Calculate both the gene-to-gene and gene-to-pathway (median or max) co-expression matrix

• python Coexpression_multiclass.py -pathway_anotation Sly_pathway_annotation_20190117_with_expression_5_members_nonoverlapping.txt -exp Results_median_FPKM_for_stress_Sly_20180125.txt -method pcc

Calculate the gene-to-gene and gene-to-pathway (median or max) mutual rank of expression similarity

• python Get_MR.py -pathway_anotation Sly_pathway_annotation_20190117_with_expression_5_members_nonoverlapping.txt -exp Results_median_FPKM_for_2017_TIBA.txt -method pcc

Data preprocessing before model building

split of genes to test, training and validation (5-fold cross-validation)

• python Split_data_five_CV.py

make features for cross-validation sets for Set B

• python Get_features_for_training_validating_testing.py path_of_expression_matrix expression_matrix number_of_cv

Get background F1s

• python Get_background_F1_for_validation_and_test.py

Naive approaches

• python Naive_prediction_crossvalidation_output_prediction.py -path ./ -expression Multiclass_MR_pcc_FC_stress.txt

Unsupervised approaches

Unsupervised approaches including: kmean (KMeans), affinity (AffinityPropagation), birch (Birch), meanshift (MeanShift)

Example for Set A

• python Clustering_crossvalidation.py -df_short_name Results_median_FPKM_for_stress_Sly_20180125.txt -path ./ -save_path ./Final_results_kmean_setA/ -clustering_method kmean -test_gene_list Genes_for_testing.txt -train_gene_list Genes_for_training.txt -dataset setA

Example for Set B

• python Clustering_crossvalidation.py -df_short_name spearman_FPKM_hormone -path ./Features_files/ -save_path ./Final_results_kmean_setB/ -clustering_method kmean -test_gene_list Genes_for_testing.txt -train_gene_list Genes_for_training.txt -dataset setB

Supervised approaches

RandomForest

• python RF_crossvalidation.py -df_short_name Results_median_FPKM_for_stress_Sly_20180125.txt -path ./ -save_path ./Final_results_RF_setA/ -test_gene_list Genes_for_testing.txt -train_gene_list Genes_for_training.txt -dataset setA

RandomForest, balance numbers of genes for each pathway for training set

• python RF_crossvalidation_SMOTE.py -df_short_name Results_median_FPKM_for_stress_Sly_20180125.txt -path ./ -save_path ./Final_results_RF_setA_SMOTE/ -test_gene_list Genes_for_testing.txt -train_gene_list Genes_for_training.txt -dataset setA

SVC, balance numbers of genes for each pathway for training set

• python SVC_SMOTE.py -df_short_name Results_median_FPKM_for_stress_Sly_20180125.txt -path ./ -save_path ./Final_results_SVC_setA_SMOTE/ -test_gene_list Genes_for_testing.txt -train_gene_list Genes_for_training.txt -dataset setA

KNeighbors, balance numbers of genes for each pathway for training set

• python KNeighbors_SMOTE.py -df_short_name Results_median_FPKM_for_stress_Sly_20180125.txt -path ./ -save_path ./Final_results_SVC_setA_SMOTE/ -test_gene_list Genes_for_testing.txt -train_gene_list Genes_for_training.txt -dataset setA

KNeighbors, balance numbers of genes for each pathway for training set

• python KNeighbors_SMOTE.py -df_short_name Results_median_FPKM_for_stress_Sly_20180125.txt -path ./ -save_path ./Final_results_KNN_setA_SMOTE/ -test_gene_list Genes_for_testing.txt -train_gene_list Genes_for_training.txt -dataset setA

DNN, balance numbers of genes for each pathway for training set

• python DNN_SMOTE.py -df_short_name Results_median_FPKM_for_stress_Sly_20180125.txt -path ./ -save_path ./Final_results_DNN_setA_SMOTE/ -test_gene_list Genes_for_testing.txt -train_gene_list Genes_for_training.txt -dataset setA

neural_network.MLPClassifier, balance numbers of genes for each pathway for training set

• python neural_network.MLPClassifier_SMOTE.py -df_short_name Results_median_FPKM_for_stress_Sly_20180125.txt -path ./ -save_path ./Final_results_MLP_setA_SMOTE/ -test_gene_list Genes_for_testing.txt -train_gene_list Genes_for_training.txt -dataset setA