Limits of Prediction for Machine Learning in Drug Discovery
Problem
Drug discovery uses ML to extrapolate molecular property predictions during optimization, but the effectiveness and limits of extrapolation had not been systematically analyzed.
Why it matters
Unreliable predictions can misdirect costly drug optimization efforts.
What the AI did
Benchmarked six ML algorithms on 243 datasets, comparing interpolation (shuffled data) with extrapolation (sorted data) using molecular weight, cLogP, and sp3-atom counts.
Results
Extrapolation with sorted data produced much larger prediction errors than shuffled data; linear ML methods were preferable for extrapolation.
Impact
Shows ML extrapolation in drug discovery carries large errors, informing model selection.
Record
| Venue | Frontiers in Pharmacology |
|---|---|
| Published | 2022 |
| Work type | article |
| Authors | Modest von Korff, Thomas Sander |
| Citations | 21 (OpenAlex, at import time) |
| References | 22 |
| Model or method | not stated in source |
| Organization | not stated in source |
| Primary source | https://doi.org/10.3389/fphar.2022.832120 |
| Open access | gold |
| Keywords | machine learning, drug discovery, extrapolation, regression, molecular properties |
Importance rating
significant — thousands to millions affected
Unreliable ML extrapolation can misdirect costly drug optimization; findings across 243 datasets inform model selection in a major industry, though no patient-level impact stated.
More in Science & Research
| # | Paper | Year | Citations | Importance |
|---|---|---|---|---|
| 1 | Accelerating drug target inhibitor discovery with a deep generative foundation model | 2023 | 51 |
8
|
| 2 | Discovery of senolytics using machine learning | 2023 | 130 |
8
|
| 3 | Deep Learning Pipeline for Accelerating Virtual Screening in Drug Discovery | 2024 | 54 |
8
|
| 4 | Matrix Factorization-based Technique for Drug Repurposing Predictions | 2020 | 56 |
8
|
| 5 | A Systematic Review of Deep Learning Methodologies Used in the Drug Discovery Process with Emphasis on In Vivo Validation | 2023 | 35 |
8
|