Fracea, Mihai-Sergiu (2026) Decoding activity in AI-redesigned enzymes Exploring Potential Activity Determinants: Integrating Sequence, Structure, and Protein Language Models. Master's Research Project 2, Biomolecular Sciences.
|
Text
mBioMol2026FraceaM.pdf Download (4MB) | Preview |
Abstract
Artificial intelligence (AI)-driven protein engineering enables exploration of sequence space beyond traditional approaches. However, there is a stringent need to understand its limitations and improve the lab validation rate. In this study, we analysed enzyme variants generated through an AI-assisted redesign pipeline, in which functional residues were fixed while other positions were mutated in silico. The variants originate from enzymes in two different subfamilies and were classified as active or inactive via lab validation. We developed a user-friendly computational pipeline integrating sequence conservation statistics, structural context, physicochemical descriptors, and protein language model embeddings to explore if some among these features can be used to predict active from inactive redesigns. XGBoost and Logistic Regression models were trained to assess the predictive potential of the features that contributed the strongest to data variation. A user-friendly tool was also used to suggest Out of Preference Mutations that should be fixed in future redesigns to refine the results of future redesigns. Our results reveal that mutations with very dissimilar residue replacement that cause large alterations in side chain sizes of conserved or buried residues are frequently associated with decreased activity. A combination of physicochemical factors and pLM embeddings separate active redesigns, however, the predictive power was enzyme specific and not transferable.
| Item Type: | Thesis (Master's Research Project 2) |
|---|---|
| Supervisor name: | Fraaije, M.W. |
| Degree programme: | Biomolecular Sciences |
| Thesis type: | Master's Research Project 2 |
| Language: | English |
| Date Deposited: | 11 Aug 2026 11:16 |
| Last Modified: | 11 Aug 2026 11:16 |
| URI: | https://fse.studenttheses.ub.rug.nl/id/eprint/37496 |
Actions (login required)
 |
View Item |