Molecular Descriptors & Fingerprints — Hands-on

Generate robust molecular features for cheminformatics and QSAR. This module covers descriptor theory and practice across 0D–3D properties and popular fingerprints, along with best practices for data hygiene, scaling, variance/collinearity diagnostics, and exportable, reproducible feature matrices.

Molecular Descriptors & Fingerprints (0D–3D, ECFP)

Help Desk · WhatsApp

Session Index

Session 1 — Descriptor Landscape (0D–3D) Session 2 — Fingerprints (ECFP/Morgan, MACCS, Paths) Session 3 — Conformers, 3D Descriptors & QA Session 4 — Mini Capstone: Feature Matrix Build

Session 1

Fee: Rs 11800

Descriptor Landscape (0D–3D)

0D/1D/2D descriptors overview

formulae, counts topological indices physicochemical props (logP, TPSA)

Tooling

RDKit Mordred PaDEL-Descriptor

Reproducibility & hygiene

standardization before calc missing/NaN handling units & provenance

Session 2

Fee: Rs 14800

Fingerprints (ECFP/Morgan, MACCS, Paths)

Binary & count fingerprints

ECFP/Morgan radius & nBits MACCS keys topological/Daylight-like

Similarity & diversity

Tanimoto/Jaccard Butina clustering scaffold splits

Sparse matrices & storage

CSR/COO formats feature hashing binary vs count use-cases

Session 3

Fee: Rs 17800

Conformers, 3D Descriptors & QA

Conformer generation

ETKDG & MMFF/UFF energy minimization ensemble selection

3D descriptors

WHIM & MoRSE autocorrelations pharmacophore counts

Quality checks

variance/near-constant filters correlation heatmaps missingness & imputation

Session 4

Fee: Rs 21800

Mini Capstone: Feature Matrix Build

Pipeline: standardize → compute → validate

Theory + Practical

Scaling & export

z-score/robust scaling sparse FP matrices CSV/NPZ + YAML config

Deliverables

feature matrix & data dictionary QA report (variance/corr) reproducible notebook/script

PDF

Molecular Descriptors & Fingerprints (0D–3D, ECFP) | RDKit, Mordred & PaDEL Hands-on

Molecular Descriptors & Fingerprints — Hands-on