Politecnico di Torino (logo)

Design and rationalizing warhead for the PROTAC targeting gamma-tubulin using neural networks

Fabiano Altieri

Design and rationalizing warhead for the PROTAC targeting gamma-tubulin using neural networks.

Rel. Jacek Adam Tuszynski, Maral Aminpour. Politecnico di Torino, Corso di laurea magistrale in Ingegneria Biomedica, 2022

PDF (Tesi_di_laurea) - Tesi
Licenza: Creative Commons Attribution Share Alike.

Download (60MB) | Preview

PROTAC technology is a new unconventional way to target and destroy tumor cells, which exploits biological processes instead of inhibiting the function of a protein, as typical of conventional drugs. Essentially, PROTAC drugs are based on 3 parts: the warhead ligand which binds to the protein of interest (POI); the E3 ligand which binds to E3 ligase and a linker, which links covalently the warhead and the E3 ligand. Moreover, PROTAC drugs aim to accelerate the degradation rate of the POIs, since they have the goal of keeping POI and E3 ligase close together for as long as possible, thus they accelerate the ubiquitination process and, consequently, the degradation mediated by proteasome. Rationalizing each part of the PROTAC is critical, but one of the great advantages is to use compounds that bind to any part of the protein and no longer just the active site, which by the way are affected by higher probability to mutate in tumor cells, effectively rendering that site untargetable. In current work, the POI is the gamma tubulin, while the E3 ligase is UBR1. Due to time constraints, it was possible focusing only on warhead, since in literature there are no known ligands binding the gamma tubulin with strong evidence. For this reason, the only alternative is virtual screening (VS) of multi-billion compounds databases such as the free access ZINC20 database. Since conventional VS (washing + conformational sampling + energy minimization + docking) is very computationally expensive even just for databases of millions compounds, DeepDock is used in order to reduce the size of ultralarge database and enrich it with the top-ranked hits, avoiding significant loss of favourable virtual hits, so that then normal docking is performed using what remains. DeepDock is a recent protocol which utilizes Quantitative structure-activity relationship (QSAR) based on deep learning models trained on docking scores of a small subset of a database to predict docking score for the rest. The models consider also the unfavorable hits, from which it draws and penalizes certain chemical groups, which in contrast in the case of docking, such hits are simply rejected, thus needlessly wasting a high amount of energy required to run docking software on HPC clusters. Additionally, molecular dynamics of gamma tubulin is performed to find the best and unique binding site among all tubulin proteins by considering electrostatic maps and similarity of residues of potential binding sites.

Relators: Jacek Adam Tuszynski, Maral Aminpour
Academic year: 2022/23
Publication type: Electronic
Number of Pages: 149
Corso di laurea: Corso di laurea magistrale in Ingegneria Biomedica
Classe di laurea: New organization > Master science > LM-21 - BIOMEDICAL ENGINEERING
Aziende collaboratrici: University of Alberta
URI: http://webthesis.biblio.polito.it/id/eprint/25789
Modify record (reserved for operators) Modify record (reserved for operators)