README for Machine Learning Model Validation Project

This repository contains the code and documentation for validating the "eos74bo" model from the Ersilia Model Hub, focusing on Tasks of the internship project, which involves checking model bias.

Project Overview

The goal of this project is to validate the accuracy and reproducibility of the "eos74bo" model, which predicts ADME properties of small molecules. Task 1 specifically involves checking for model bias by running predictions for a list of 1000 diverse molecules and plotting the results in a scatter plot. and Task 2 reproducibity of results and Task 3 is extrnal data validation from Selected paper

About Selected model

Selected model

Kinetic aqueous solubility (μg/mL) was experimentally determined using the same SOP in over 200 NCATS drug discovery projects. A final dataset of 11780 non-redundant molecules and their associated solubility was used to train a SVM classifier. Approximately half of the dataset has poor solubility (< 10 μg/mL), and two-thirds of these low soluble molecules report values of < 1 μg/mL. A subset of the data used is available at PubChem (AID 1645848) https://pubchem.ncbi.nlm.nih.gov/bioassay/1645848#section=Result-Definitions.

Characteristics

• Input: Compound
• Input Shape: Single
• Task: Classification
• Output: Probability
• Output Type: Float
• Output Shape: Single
• Interpretation: Probability of a compound having poor solublibity (< 10 µg/ml)

Repository Structure

• README.md: This file, providing an overview of the project and instructions for reproducing the work.
• src/: Directory containing source code.
• data/: Directory containing datasets used for validation.
• results/: Directory containing output plots and any other result files.
• LICENSE: License file for the repository.
• .gitignore: File specifying intentionally untracked files to ignore.

Getting Started

Task 1 - Week 2

Evaluating the biased

• Have used the project from the mentioned models
• Dowloaded the dataset
• Cleaned the dataset
• Converted the smiles to standard
• Generated the InchyKeys
• Made the predictions
• Ploted the scttered plot

Task 2 - Week 2

Reproducibility of Results

Select the paper and get the dataset that used in that paper and make prediction on that.

The Paper is followed:

https://slas-discovery.org/action/showPdf?pii=S2472-5552%2822%2906765-X that used the Solubilty dataset:

About the Code:

In this notebook, I am loading a list of molecules I obtained from PubChem for solubilty check of drug, and will replicate the results for Ersilia Hub Model mentioned in link and for the git hub code of external sourse mentioned in linkprocessing them to make sure I have:

• Some data visulization
• Generate the ROC curve for the original github code
• Generate the ROC curves for the EMH code
  • Outcome generated after standardized smiles
• Make the comparison

OutCome:

• Is about the substance id highly Soluble or active class = '0' or Low Soluble or Inactive class = '1'
• Outcome is in the probability

Github model Run

App running in Terminl

Download the Dataset for prediction

Upload the Dataset for predictions

Task 3 - Week 3

PAMPA, parallel artificial membrane permeability assay: PAMPA is a laboratory test used in drug development to assess how easily a drug can pass through cell membranes. It helps researchers understand the ability of a drug to be absorbed into the bloodstream, which is important for determining its effectiveness.

Download the External dataset:

the code to download the dataset is below

from tdc.single_pred import ADME
data = ADME(name = 'PAMPA_NCATS')
split = data.get_split()