| Titre : | An AI tool for drug discovery and repurposing: Molecular Property Prediction |
| Auteurs : | Asma Djennad, Auteur ; Reguia Kherbich, Auteur ; Belkacem Abdelli, Auteur |
| Type de document : | Mémoire magistere |
| Editeur : | Biskra [Algérie] : Faculté des Sciences Exactes et des Sciences de la Nature et de la Vie, Université Mohamed Khider, 2024 |
| Format : | 1 vol. (77 p.) / ill.couv.ill.encoul / 30cm |
| Langues: | Anglais |
| Langues originales: | Anglais |
| Résumé : |
Drug discovery is considered as a complex and time-consuming process which involves design,development and testing new drug candidates for the treatment of numerous diseases. Over thepast decade, artificial intelligence (AI) has transformed drug discovery and development practices,enabling it to intervene at various stages. Among these stages is hit identification, where theactual search for drug candidates to treat particular disease begins. Once drug candidates are identified, their efficacy must be evaluated to determine their potential as drug candidates. This evaluation is guided by molecular property prediction (MPP) task. In recent years, advancements in AI have enhanced MPP by providing a powerful tools in order to accelerate drug discovery. Inthis work, we propose to develop a pretraining model called SELF-BERT for SELFIES sequences,based on the BERT Transformer model to predict ADMET molecular properties (Absorption,Distribution, Metabolism, Excretion, Toxicity). The model was later integrated into a user-friendlyplatform (Algeria SmartPharma) to facilitate the visualization of predicted results. The obtained results indicate that our model achieves an accuracy of 96% during the pretraining phase anddemonstrates robust and comparable performance in the fin-tuning phase against state-of-the-art models in classification and regression tasks. |
| Sommaire : |
AcknowledgementsConferences and animationsAbstractList of figuresList of tables General introduction 1 1 Pharmaceutical and AI Background 4 1.1 Introduction . . . 4 1.2 Pharmaceutical industry . . . . . . 4 1.2.1 Primary actors of pharmaceutical industry . . . . . . . . . . . . . . . 5 1.3 Drug discovery and development . . . . 6 1.3.1 Drug discovery and development process . . . . . . . . . . . . . . . . . . . . 6 1.3.1.1 Drug discovery process . . . . . . . . . . . . . . . 6 1.3.1.2 Drug development process . . . . . . . . . . . . . 8 1.3.2 Molecular fundamentals . . . . . . . . . . . . . . . . . . . . 10 1.3.2.1 What is molecule? . . . . . . . . . . . . . . . . 10 1.3.2.2 Molecular representations . . . . . .. . . . . . . . 10 1.3.2.3 Molecular properties . . . . . .. . . . . . . . 12 1.3.2.4 ADMET properties . . . . . . . . . . . .. . . . . . 13 1.4 Artificial intelligence . . . . . 15 1.4.1 Machine learning . . . 16 1.4.1.1 Types of learning . . . . . . . . . . . . . . . . . . . 16 1.4.2 Deep learning . . . . 17 1.4.2.1 Types of DL models . . . . . . . . . . . . . . . . . . . . . . . . . . 17 1.5 AI in drug discovery and development .22 1.5.1 AI in drug discovery process . . 23 1.5.2 AI in drug development process . . . . . . . . . . . . . . . . . . . . . . . . . 24 1.5.3 Related works about AI in Molecular Property Prediction . . . . . . . . . . 24 1.5.4 Discussion . . .28 1.5.5 Motivation . . . .29 1.5.6 Proposed method . . . . . . . . . . . . 29 1.6 Conclusion . . . . 30 2 System design and implementation 31 2.1 Inroduction . . . . . . . 31 2.2 Work objectives . . . . . 31 2.3 Global system architecture . . . 32 2.3.1 Data layer . . . . . 32 2.3.2 Model layer . . . .33 2.3.3 End-User layer . . . . . . . . 33 2.4 Detailed Model layer architecture . . . . 34 2.4.1 Preparation phase . . 34 2.4.2 Pretraining phase . . .35 2.4.3 Fin-tuning phase . . . . .36 2.5 End-user layer analysis and design . . . . 38 2.5.1 Functional requirements: Use Case Diagram . . . . . . . . . . . . . . . . . . 38 2.5.2 Non-functional requirements . . . . . . . . . . . . . . . . . . . . . . . . . . . 41 2.5.3 Static aspect: Class Diagram . . . . . . . . . . . . . . . . . . . . . . . . . . 41 2.5.4 Dynamic aspect: Sequence Diagram . . . . . . . . . . . . . . . . . . . . . . . 43 2.6 Performance evaluation metrics . . . 45 2.6.1 Classification performance metrics . . .45 2.6.2 Regression performance metrics . . . . . . . . . . . . . . . . . . . . . . . . . 46 2.7 Implementation . . . . . 47 2.7.1 Configuration of materials used . . . . . . . . . . . . . . . . . . . . . . . . . 47 2.7.1.1 Materials used for AI model development . . . . . . . . . . . . . . 47 2.7.1.2 Materials used for platform development . . . . . . . . . . . . . . . 47 2.7.2 Development tools and technologies . . . . . . . . . . . . . . . . . . . . . . . 48 2.7.2.1 Development environments . . . . . . . . . . . . . . . . . . . . . . 48 2.7.2.2 Programming languages . . . . . . . . . . . . . . . . . . . . . . . . 49 2.7.2.3 Key libraries . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 50 2.7.3 AI model development . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51 2.7.3.1 Data preparation step . . . . . . . . . . . . . . . . . . . . . . . . . 51 2.7.3.2 Pretraining implementation . . . . . . . . . . . . . . . . . . . . . . 53 2.7.3.3 Fin-tuning implementation . . . . . . . . . . . . . . . . . . . . . . 54 2.7.4 Integrating the model into the platform . . . . . . . . . . . . . . . . . . . . . 56 2.8 Conclusion . . . 57 3 Experimentation results and discussion 58 3.1 Inroduction . . . . . . . 58 3.2 Presentation of the Algeria SmartPharma platform . . . . . 58 3.3 Results of AI model . . . . . . 63 3.3.1 Pretraining phase . . . .63 3.3.2 Fin-tuning phase . . . . . .64 3.4 Comparison with existing research . . .. 65 3.4.1 Pretraining phase . . .65 3.4.2 Fin-tuning phase . . . 66 3.5 Discussion . . . . 70 3.5.1 Pretraining performance . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 70 3.5.2 Molecular property prediction performance . . . . . . . . . . . . . . . . . . . 71 3.6 Comparison with existing platforms . . . 71 3.7 Case Study: Marburg Virus . . . . . 73 3.7.1 What is Marburg Virus? . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 73 3.7.2 Virology and molecular structure of MARV . . . . . . . . . . . . . . . . . . 74 3.7.3 Drug discovery efforts against MARV . . . . . . . . . . . . . . . . . . . . . . 74 3.7.4 Molecular Property Prediction for MARV drug candidates . . . . . . . . . . 76 3.8 Conclusion . . . . . . 77 Conclusion and Perspectives 78 Bibliography79 |
| Type de document : | Mémoire master |
Disponibilité (1)
| Cote | Support | Localisation | Statut |
|---|---|---|---|
| MINF/929 | Mémoire master | bibliothèque sciences exactes | Consultable |
