Quantitative interpretation explains machine learning models for chemical reaction prediction and uncovers bias

Dávid Péter Kovács, William McCorkindale, Alpha A. Lee

Nature Communications 2021 article cited by 114

Source Open-access PDF More in Science & Research

importance 6/10 ai assisted Computational Drug Discovery Methods drug discovery stage: research

Problem

Machine-learning models for organic reaction prediction are opaque black boxes, so it is unclear whether predictions reflect real chemistry or dataset bias.

Why it matters

Organic synthesis is a major bottleneck in drug discovery, and untrustworthy model predictions hinder its use.

What the AI did

Built a quantitative interpretation framework attributing Molecular Transformer predictions to reactant parts and training reactions, retrieving evidence and uncovering 'Clever Hans' bias-driven predictions.

Results

not stated in source

Impact

Evidence shows the framework explains predictions and identifies dataset bias; no numeric benchmarks stated.

Record

Venue	Nature Communications
Published	2021
Work type	article
Authors	Dávid Péter Kovács, William McCorkindale, Alpha A. Lee
Citations	114 (OpenAlex, at import time)
References	47
Model or method	Molecular Transformer
Organization	not stated in source
Primary source	https://doi.org/10.1038/s41467-021-21895-w
Open access	gold
Keywords	machine learning interpretability, reaction prediction, Molecular Transformer, Clever Hans, dataset bias

Importance rating

6 / 10

significant — thousands to millions affected

Organic synthesis is a major drug-discovery bottleneck per abstract; interpretability framework builds trust in reaction prediction, but no scale or adoption figures are stated.

Basis: inferred. Judged from the source text, which gives no explicit scale. Stated scale: not stated in source. Method and limits.

More in Science & Research

#	Paper	Year	Citations	Importance
1	Accelerating drug target inhibitor discovery with a deep generative foundation model Vijil Chenthamarakshan, Samuel C. Hoffman, David Owen, Petra Lukacik,...	2023	51	8
2	Discovery of senolytics using machine learning Vanessa Smer-Barreto, Andrea Quintanilla, Richard J. R. Elliott, John...	2023	130	8
3	Deep Learning Pipeline for Accelerating Virtual Screening in Drug Discovery Fatima Noor, Muhammad Junaid, Atiah H. Almalki, Mohammed Almaghrabi, S...	2024	54	8
4	Matrix Factorization-based Technique for Drug Repurposing Predictions Gaia Ceddia, Pietro Pinoli, Stefano Ceri, Marco Masseroli	2020	56	8
5	A Systematic Review of Deep Learning Methodologies Used in the Drug Discovery Process with Emphasis on In Vivo Validation Nikoletta-Maria Koutroumpa, Konstantinos D. Papavasileiou, Anastasios...	2023	35	8