app.py

import gradio as gr
import pickle

gender_categories = ["Male", "Female"]
gender = gr.inputs.Dropdown(gender_categories, label="Gender")

race_categories = ["White", "Black", "Other"]
race = gr.inputs.Dropdown(race_categories, label="Race")

age_categories = ["80 - 89", "70 - 79", "50 - 59", "60 - 69",  "0 - 9", "40 - 49",  "30 - 39", "20 - 29", "10 - 19"]
age =  gr.inputs.Dropdown(age_categories, label="Age")

height = gr.inputs.Textbox(label = "Height (cm)")
weight =  gr.inputs.Textbox(label = "Weight (kg)")

diabetes_categories = ["Yes", "No"]
diabetes = gr.inputs.Dropdown(diabetes_categories, label="Diabetes")

simvastatin_categories = ["Yes", "No"]
simvastatin = gr.inputs.Dropdown(simvastatin_categories, label="Simvastatin (Zocor)")

amiodarone_categories = ["Yes", "No"]
amiodarone  = gr.inputs.Dropdown(amiodarone_categories, label="Amiodarone (Cordarone)")

target_inr = gr.inputs.Textbox(label = "Target INR (mg/week)")
inr = gr.inputs.Textbox(label = "INR on Reported Therapeutic Dose of Warfarin")

cyp2c9_genotypes_categories = ["*1/*1", "*1/*3", "*1/*2", "*2/*2", "*1/*5", "*2/*3", "*3/*3"]
cyp2c9_genotypes = gr.inputs.Dropdown(cyp2c9_genotypes_categories, label = "Cyp2C9 genotypes")

vk0rc1_genotype_categories = ["A/G", "G/G", "A/A"]
vk0rc1_genotype = gr.inputs.Dropdown(vk0rc1_genotype_categories, label = "VKORC1 genotype: -1639 G>A (3673); chr16:31015190; rs9923231; C/T")

model_selection = gr.inputs.Dropdown(["Linear Regression (Regression)", "Ridge Regression (Regression)", "Random Forest (Regression)", "Neural Network (Regression)", "K-Nearest Neighbors (Classification)", "Logistic Regression (Classification)", "Random Forest (Classification)"], label="Machine Learning Model")

warfarin_dose = gr.outputs.Textbox(label = "Warfarin Dose")

linear_model = pickle.load(open('linear_model_regression.pkl', 'rb'))
ridge_model = pickle.load(open('ridge_model_regression.pkl', 'rb'))
forest_mode = pickle.load(open('forest_model_regression.pkl', 'rb'))
ann_model = pickle.load(open('ann_model_regression.pkl', 'rb'))
knn_model_clf = pickle.load(open('knn_model_classification.pkl', 'rb'))
logistic_model_clf = pickle.load(open('logistic_model_classification.pkl', 'rb'))
random_forest_model_clf = pickle.load(open('random_forest_model_classification.pkl', 'rb'))

def predict_warfarin_dose(gender, race, age, height, weight, diabetes, simvastatin, amiodarone, target_inr, inr, cyp2c9_genotypes, vk0rc1_genotyp, model_selection):
    new_instance = [gender_categories.index(gender), race_categories.index(race), age_categories.index(age), height, weight, diabetes_categories.index(diabetes), simvastatin_categories.index(simvastatin), amiodarone_categories.index(amiodarone), target_inr, inr, cyp2c9_genotypes_categories.index(cyp2c9_genotypes), vk0rc1_genotype_categories.index(vk0rc1_genotype)]
    if model_selection == "Linear Regression (Regression)":
        prediction = linear_model.predict(new_instance)
    elif model_selection == "Ridge Regression (Regression)":
        prediction = ridge_model.predict(new_instance)
    elif model_selection == "Random Forest (Regression)":
        prediction = forest_mode.predict(new_instance)
    elif model_selection == "Neural Network (Regression)":
        prediction = ann_model.predict(new_instance)
    elif model_selection == "K-Nearest Neighbors (Classification)":
        prediction = knn_model_clf.predict(new_instance)
    elif model_selection == "Logistic Regression (Classification)":
        prediction = logistic_model_clf.predict(new_instance)
    else:
        prediction = random_forest_model_clf.predict(new_instance)
    return prediction

interface = gr.Interface(fn=predict_warfarin_dose, inputs=[gender, race, age, height, weight, diabetes, simvastatin, amiodarone, target_inr, inr, cyp2c9_genotypes, vk0rc1_genotype, model_selection], outputs=warfarin_dose)
interface.launch()