Descriptor Engineering & Feature Selection — Hands-on

Learn how to turn raw chemical structures into clean, informative feature matrices ready for QSAR, ML and property prediction. This module walks through molecular descriptor families, descriptor calculation, quality control, scaling and leakage-aware feature selection so that downstream models remain stable, interpretable and reproducible.

Descriptor Engineering & Feature Selection

Help Desk · WhatsApp

Session Index

Session 1 — Molecular Descriptors Landscape (0D–3D) Session 2 — Descriptor Calculation & Data Hygiene Session 3 — Feature Selection, Reduction & Leakage Checks Session 4 — Mini Capstone: QSAR Ready Feature Matrix

Session 1

Fee: Rs 8800

Molecular Descriptors Landscape (0D–3D)

Descriptor families and levels of representation

0D / 1D / 2D / 3D descriptors global vs local features scalar vs vector forms

Physicochemical and topological descriptors

MW, logP, pKa, polar surface topological indices fragment counts

3D and conformation dependent descriptors

3D pharmacophoric patterns shape descriptors when 3D really matters

Session 2

Fee: Rs 11800

Descriptor Calculation & Data Hygiene

Descriptor engines and toolkits overview

RDKit and mordred PaDEL and CDK concepts batch processing of libraries

Handling missing values and unstable descriptors

constant and near constant columns error flags and NaNs removal vs imputation

Scaling, normalization and transformation choices

standard vs min max scaling log transform and power methods robust scaling for outliers

Session 3

Fee: Rs 14800

Feature Selection, Reduction & Leakage Checks

Multicollinearity diagnostics and pruning

correlation heatmaps variance inflation factor redundant feature removal

Filter, wrapper and embedded methods

univariate filters and mutual information forward and backward selection Lasso, tree based importance

Dimensionality reduction and leakage aware workflows

PCA and latent spaces pipelines in scikit learn train only fitting and data leakage

Session 4

Fee: Rs 18800

Mini Capstone: QSAR Ready Feature Matrix

From structures to curated descriptor table

Theory + Practical

End to end feature engineering and selection workflow

descriptor generation script quality checks and pruning split aware pipelines

Deliverables: feature matrix, code and documentation

CSV feature matrix notebook or script file readme for QSAR teams

PDF

Descriptor Engineering & Feature Selection Training | Molecular Descriptors & QSAR Feature Space

Descriptor Engineering & Feature Selection — Hands-on