CHEM 487/542 – Chemical Data Science

This course explores advanced topics in data science and artificial intelligence (AI) with a focus on chemistry applications. It covers the fundamentals of machine learning and data science, along with advanced research areas such as ML-guided experimental design, high-throughput screening, and AI-driven chemical discovery. Through a combination of lectures and hands-on lab sessions, students will gain both the theoretical background and practical skills needed to pursue research in chemical data science.

Evaluation:

– Class participation (20%): Includes attending lectures and hands-on sessions and actively asking questions.

– Hands-on assignments (50%): Hands-on assignments include coding and computational tasks, along with analysis of the results.

– Capstone project (30%): A short research-related project that will be summarized as a project report and a 20-minute talk.

Schedule and Course Materials

Lecture notes and lab codes for the first 17 lectures were created by Dr. Chong Sun. The English was refined with assistance from ChatGPT. More lectures will be added, stay tuned!

All rights reserved. If you use any part of this content, please provide proper acknowledgment of the source. If you enjoy the course, please give a star to the Course Github Repo!

#	Date	Lecture	Lab
1.	Jan 20	Introduction: Data science for Chemistry	1. Python Basics 2. UV-Vis Spectrum
2	Jan 22	Chemical data: Representation	1. 1D string and 3D XYZ 2. Graph representation.
3	Jan 27	Chemical data: Acquisition and Feature Engineering	1. API Access 2. Chemical Similarity
4	Jan 29	Fundamentals of Statistical Analysis	Maximum Likelihood Estimation
5	Feb 3	Statistical Methods for Chemical Prediction	1. Solubility prediction with linear models 2. Boiling point prediction with Gaussian process
6	Feb 5	Dimension reduction and Preliminaries on Training ML Models	Feature selection and PCA
7	Feb 10	Feedforward Neural Networks	1. Build an FNN with PyTorch 2. Use GPU backend
8	Feb 12	Chemical Feature Engineering II	1. Ultrafast shape recognition 2. Simple learned embedding for molecules
9	Feb 17	Case Study: Learning a potential energy surface (PES)	1. MLP with ACSF and FNN 2. MLP with SOAP and GPR
10	Feb 19	Chemical Prediction with Graph neural networks (GNN)	1. Molecular property prediction with GNN 2. Create a Graph Dataset
11	Feb 24	Active Learning for Chemical Data Efficiency	1. Improve GPR with active learning 2. Improve FNN with MC dropout and max-min sampling
12	Feb 26	Chemical Design with Bayesian optimization	BH Reaction optimized by BO
13	Mar 3	Reaction kinetics modeling with recurrent neural networks (RNN)	1. Reaction Kinetics with RNN 2. Molecular Generation with RNN (LSTM)
14	Mar 5	Retrosynthesis with Sequence-to-Sequence Modeling	Retrosynthesis with Seq2Seq model
15	Mar 10	Generative Chemistry and Variational Autoencoder	Learning the Chemical Latent Space with VAE
16	Mar 12	Molecular Docking with Diffusion Models	1. Processing a PDB file 2. A simple implementation of Diffusion Models
17	Mar 24	Crystal Structure Design with Transformer
18	Mar 26	Equivariance Neural Networks	Laurence Giordano
19	Mar 31	Self-Driving Lab and Agentic AI	Self-reading
21	Apr 7	Help session with terminal, WSL, VSCode, Git/Github
22	Apr 9	Guest lecture (Prof. York)
23	Apr 16	Guest lecture (Prof. Khare)
24	Apr 23	Guest lecture (Prof. Remsing)