Step 1: Input file and options

1. You have the option to select your own file in .csv, tab-delimited, excel and RDS format, or use one of the example data set, Nilsson rare, Mosmann rare, and SIN3 network.

2. Select your categorical response variable (formatted as 0 and 1) from “Choose binary outcome”.

3. Select the variables that you don’t want in your data analysis from the “Select unwanted variables”.

Step 2: Resampling Methods

You have the option to select one of the two resampling techniques for class imbalance from the “Resampling Methods” option:

1. SMOTE
2. ADASYN
3. None

Please select “None” when you don’t want to use resampling methods in your data analysis.

Step 3: Normalization

You have the option to select one of the following normalization techniques from the “Normalization Technique” option:

1. Min-max
2. Log Scaling
3. Standard Scaling
4. Arcsine
5. TopS
6. Percentage Row
7. No Normalization

Please select “No Normalization” when you don’t want to to normalize your data or have already performed normalization outside PerSEveML.

Step 4: Machine Learning Models

You can select one or more machine learning (ML) algorithms based on your research question from the “Choose preferred algorithms” option. The options include:

• Tree-based algorithms:

1. Decision Tree
2. Random Forest
3. XgBoost
4. AdaBoost

• Non-tree based algorithms:

1. Naive Bayes
2. Linear SVM
3. Non-linear SVM
4. Polynomial SVM

• Linear classifiers:

1. LDA
2. Logistic Regression
3. Lasso Regression
4. Ridge Regression

Step 5: Select the train-test percentage

The user can select can either enter or select any percentage between 0-100 to split the normalized/raw data into training and test sets using the “Insert train-test ratio” tab. Note that test set should have at least some data points for the app to run successfully.

Step 6: Select the value of k for cross-validation

This app heavily rely on the performance of the ML methods. Since the algorithm is based on hyper-parameter tuning using grid search or cross-validation, using an optimum value of k is crucial. This app allows the user to select the value of k between 1 to 10 in the “Value of k for cross validation” tab.

Step 7: Select the threshold for cutpoint analysis

Based on the user defined cut-point in the “Insert cut-point analysis cutoff”, this app will formulate the persistent biomarker (or feature) structure. An user can use different combinations of normalizations, ML methods, and thresholds to come up with the most peristent structure.

Step 8: Running the app

After the user have successfullly selected the preferred input options, they can go ahead and click on the gree “Submit” button of the upper left corner.