Molecular recognition of protein-ligand complexes via convolutional neural networks

dc.contributor.author Güner, Hüseyin
dc.contributor.department AGÜ, Fen Bilimleri Enstitüsü, Elektrik ve Bilgisayar Mühendisliği Ana Bilim Dalı en_US
dc.date.accessioned 2022-09-01T11:13:29Z
dc.date.available 2022-09-01T11:13:29Z
dc.date.issued 2022 en_US
dc.date.submitted 2022-01
dc.description.abstract As a sub-discipline of Artificial Intelligence, deep neural networks have received enormous interest in research and industrial applications over the last decades owing to their highly successful performance in addressing and solving broad areas of problems. Hence, especially hitherto achievements in computer-aided drug design brought an extra impetus with the novel deep learning approaches in structure-based drug design etiology. Our group offers a novel convolutional neural network model, deepMLR, that casts insight into the molecular recognition of ligand molecules and a receptor protein molecule. Having compared our model and a few other existing models with a case study of a traditional approach, herein, we present the success story of a deep learning model straight. en_US
dc.description.abstract Yapay Zeka'nın bir alt disiplini olarak derin sinir ağları, geniş spektrumdaki problem alanlarını ele alma ve çözmedeki son derece başarılı performansları nedeniyle, son on yılda (özellikle) araştırma ve endüstriyel uygulamalarda büyük bir ilgi görmeye başladı. Özellikle son zamanlardaki, bilgisayar destekli ilaç tasarımındaki başarıları nedeniyle, yapı tabanlı ilaç tasarımı etiyolojislerindeki yeni derin öğrenme yaklaşımlarına karşı ekstra bir ivme kazanmıştır. Grubumuz, ligand moleküllerinin ve bir reseptör protein molekülünün moleküler olarak tanınması hakkında bir fikir veren yeni bir konvolüsyonel sinir ağı modeli sunmaktadır. Diğer mevcut modellerle ve modelimizle geleneksel bir yaklaşımın örnek çalışmasıyla karşılaştırıldığında, burada derin bir öğrenme modelinin başarı hikayesini sunuyoruz. en_US
dc.description.tableofcontents 1. INTRODUCTION .................................................................................................... 1 2. STRUCTURE BASED DRUG DISCOVERY........................................................ 3 2.1 MOLECULAR RECOGNITION ..................................................................................... 5 2.1.1 Thermodynamic Entities.................................................................................. 6 2.2 LIGAND-PROTEIN AFFINITY SCORING FUNCTIONS................................................... 7 2.3 MOLECULAR DYNAMICS SIMULATIONS ................................................................... 8 2.3.1 Force fields in protein-ligand complexes ........................................................ 9 2.4 THE MM/PBSA AND MM/GBSA METHODS............................................................ 9 3. DEEP LEARNING METHODS............................................................................ 11 3.1 ARTIFICIAL INTELLIGENCE IN DRUG DISCOVERY................................................... 11 3.1.1 Feed-forward neural networks ...................................................................... 12 3.1.2 Convolutional Neural Networks .................................................................... 16 4. EXPERIMENTAL RESULTS............................................................................... 20 4.1 COMPUTATIONAL RESOURCES ............................................................................... 20 4.2 CONVENTIONAL/TRADITIONAL SBDD CAMPAIGN ................................................ 20 4.2.1 Structural Data Files and Preparation for Docking Experiments ................ 20 4.2.2 Docking experiments...................................................................................... 22 4.2.3 MD Experiments............................................................................................ 24 4.3 A 3D CNN MODEL BY PAFNUCY............................................................................ 25 4.3.1 Datasets ......................................................................................................... 26 4.3.2 Architecture of the network............................................................................ 27 4.3.3 Training with Back-propagation ................................................................... 27 4.3.4 Results and metrics of back-propagation ...................................................... 28 4.3.5 Binding affinity of 2SHP with the leads......................................................... 30 4.4 OUR MODEL DEEPMLR......................................................................................... 30 4.4.1 Architecture of DeepMLR.............................................................................. 32 4.4.2 Training and evaluation of DeepMLR........................................................... 34 4.4.3 Evaluation Metrics......................................................................................... 35 4.4.4 Results of experiments run by DeepMLR....................................................... 35 5. CONCLUSIONS AND FUTURE PROSPECTS ................................................. 40 5.1 CONCLUSIONS ........................................................................................................ 40 5.2 SOCIETAL IMPACT AND CONTRIBUTION TO GLOBAL SUSTAINABILITY................... 41 5.3 FUTURE PROSPECTS ............................................................................................... 42 6. BIBLIOGRAPHY................................................................................................... 43 en_US
dc.identifier.uri https://hdl.handle.net/20.500.12573/1366
dc.language.iso eng en_US
dc.publisher Abdullah Gül Üniversitesi Fen Bilimleri Enstitüsü en_US
dc.relation.publicationcategory Tez en_US
dc.rights info:eu-repo/semantics/openAccess en_US
dc.subject Molecular Recognition en_US
dc.subject Structure Based Drug Design en_US
dc.subject Deep Convolutional Neural Networks en_US
dc.subject Protein-Ligand Affinity Prediction en_US
dc.subject Structure Based Virtual Screening en_US
dc.title Molecular recognition of protein-ligand complexes via convolutional neural networks en_US
dc.title.alternative Protein-ligand komplekslerinin konvolüsyenel sinir ağları ile moleküler tanınması en_US
dc.type masterThesis en_US

Files

Original bundle

Now showing 1 - 1 of 1
Loading...
Thumbnail Image
Name:
714262_hüseyingüner.pdf
Size:
3.02 MB
Format:
Adobe Portable Document Format
Description:
tez

License bundle

Now showing 1 - 1 of 1
No Thumbnail Available
Name:
license.txt
Size:
1.44 KB
Format:
Item-specific license agreed upon to submission
Description: