Klasifikátor vazebných kapes v proteinech založený na strojovém učení
| Title in English | A machine learning-based classifier of binding pockets in proteins |
|---|---|
| Authors | |
| Year of publication | 2024 |
| Type | Software |
| MU Faculty or unit | |
| web | Je k dispozici jako supplement rukopisu |
| Description | Certain structural elements of proteins play a critical role in binding and unbinding ligands, e.g., the surface binding interface, a tunnel with one opening that connects the active site with the outside environment, or a channel open on both sides. There are several methods for tunnel and channel calculation, but they require customized settings to produce high-quality results. In particular, the discrimination between surface and buried binding pockets is critical for tunnel calculation but is an open problem. This classifier aims to solve this problem using FPOCKET features and manually labelled dataset of 200 pockets classified as surface, borderline, and buried. The classifier consists of a small artificial neural network and achieves the accuracy of 54% and F1 score of 50%. |
| Related projects: |