Cheminformatics Tutorial

Version 1.0 - June, 2026. Monterrey

---

Description

This tutorial on Cheminformatics in Drug Discovery simulates a six-month research workflow, guiding the transition from raw chemical data acquisition to predictive modeling and ligand-based virtual screening of large chemical libraries.

---

Pedagogical Approach

This tutorial follows a bottom-up computational workflow, progressively constructing the pipeline from raw molecular representations to predictive modeling and ligand-based virtual screening.

Pedagogically, it adopts a project-based learning (PBL) framework. Learners simulate the responsibilities of a computational chemist within a structured research timeline, developing technical rigor, methodological coherence, and reproducible scientific practice.

---

Curriculum Overview

MODULE 0: Environment Setup

Focus: Preparing a reproducible computational environment for cheminformatics workflows.

• Platform Setup: Configuration of a reproducible working environment using Google Colab, Jupyter, Anaconda, and VS Code.

MODULE 1: Molecular Data Acquisition and Representation

Focus: Establishing the digital foundation of chemical matter.

• Programmatic Access: Automated data retrieval from PubChem and ChEMBL using Python APIs.
• Chemical Notation: Introduction to SMILES, InChI, InChIKey, and common structure file formats sucha as SDF and MOL.
• Molecular Representation: Conversion of chemical encodings into computational molecular objects for downstream analysis.
• Chemical Visualization: Generation of 2D molecular depictions for structural inspection and interpretation.

MODULE 2: Data Curation & Structural Exploration

Focus: Preparing chemically consistent datasets for analysis and modeling.

• Data Curation: Structural standardization, salt stripping, tautomer handling, stereochemistry normalization, and duplicate removal using RDKit.
• Structure-level EDA: Inspection of duplicate entries, scaffold diversity, atom composition, molecular size trends, and class balance.
• Chemical Quality Control: Identification of inconsistent structures, missing values, and dataset-level issues that may affect downstream analysis.
• Preliminary Chemical Space Inspection: Early visualization of structural diversity and dataset composition prior to descriptor generation.

MODULE 3: Molecular Descriptors, Fingerprints & Feature Engineering

Focus: Transforming chemical structures into machine-readable features.

• Molecular Descriptors: Calculation of key 1D and 2D physicochemical properties, including molecular weight (MW), LogP, TPSA, HBD, HBA, and rotatable bonds.
• Descriptor-Based EDA: Distribution analysis of calculated properties, correlation assessment, outlier detection, and redundancy evaluation.
• Molecular Fingerprints: Generation of structural fingerprints for similarity analysis, diversity assessment, and machine learning applications.
• Feature Engineering: Preparation and organization of descriptor matrices and fingerprint-based representations for predictive modeling.
• Chemical Space Visualization: Use of methods such as PCA for preliminary visualization of descriptor-based chemical space.

MODULE 4: Predictive Modeling

Focus: Learning structure–activity relationships through machine learning.

• Model Inputs: Use of molecular descriptors and fingerprint matrices as representations for classification tasks.
• Model Training: Supervised learning algorithms such as Logistic Regression and Random Forest for bioactivity prediction.
• Model Validation: Evaluation using ROC-AUC, precision-recall analysis, and cross-validation strategies.
• Performance Interpretation: Analysis of model behavior, feature importance, and predictive limitations.
• Structure–Activity Learning: Introduction to computational strategies for relating molecular features to biological activity outcomes.

MODULE 5: Virtual Screening & Scientific Implementation

Focus: Applying predictive models to prioritize candidate compounds.

• Ligand-Based Virtual Screening: Application of trained models to large chemical libraries such as COCONUT or selected ChEMBL subsets.
• Candidate Prioritization: Probability thresholding, ranking, and filtering strategies for hit selection.
• Result Interpretation: Evaluation of screened compounds in terms of confidence, diversity, and practical relevance.
• Scientific Reporting: Preparation of technical summaries, screening reports, and structured outputs for scientific communication.

---

Purpose

The purpose of this tutorial is to help you:

• Master Molecular Representation: Develop a rigorous understanding of how chemical structures are digitized, standardized, and manipulated using computational tools.

• Implement Automated Workflows: Transition from manual data handling to automated, programmatic retrieval and curation of chemical data from global databases such as ChEMBL and PubChem.

• Integrate Chemistry and Machine Learning: Apply statistical and machine learning methods to model structure–activity relationships and predict biological activity..

• Perform Ligand-Based Virtual Screening: Deploy trained models to evaluate and prioritize compounds from large chemical libraries (e.g., COCONUT).

• Develop Professional Computational Practices: Simulate the workflow of a computational chemist, emphasizing reproducible code, rigorous validation, and clear scientific reporting..

---

Repository Structure

├── 00_Environment_Setup/       # Installation
├── 01_Data_Acquisition/        # APIs, SMILES, InCHI, SDF/MOL 
├── 02_Data_Curation/           # Data Curation, structural EDA
├── 03_Feature_Engineering/     # Fingerprints, descriptor-based EDA
├── 04_Predictive_Modeling/     # Machine Learning, validation, prediction
├── 05_Virtual_Screening/       # Ligand-based screening, identification, priorization
└── 06_Reporting/               # Technical reports

---

Requirements

• Anaconda
• Python 3.10+
• RDKit (Cheminformatics engine)
• Pandas / NumPy (Data manipulation)
• Scikit-Learn (Machine Learning)
• Matplotlib / Seaborn (Visualization)
• VS Code (IDE)
• MiKTex-XeTeX (Documentation)
• environment.yml

---

Authors: Flavio F. Contreras-Torres. Tecnológico de Monterrey.
Monterrey, Mexico

Contributors: Fernanda I. Saldivar-González. Tecnológico de Monterrey.
Monterrey, Mexico

---

Versions

v.1.0 - June 2026

---

License

The content of this tutorial itself is licensed under the terms and conditions of the Creative Commons Attribution (CC BY 4.0) license, and the underlying source code used to format and display that content is licensed under the MIT License. See the LICENSE file for full details.

Attribution

If you use or adapt this material, please provide appropriate credit to the original authors and contributors, and repository:

NanoBiostructures Research Group
GitHub: https://github.com/NanoBiostructuresRG