Artificial Neural Network for Drug Design, Delivery and Disposition

Artificial Neural Network for Drug Design, Delivery and Disposition

Editors

Munish Puri

Yashwant Pathak

Vijay Kumar Sutariya

Srinivas Tipparaju

Wilfrido Moreno

Table of Contents

Cover image

Title page

Copyright

Dedication

Contributors

Foreword

Preface

Section 1. Basics of ANN: Concept and Strategy in Drug Design

Chapter 1. Introduction to Artificial Neural Network (ANN) as a Predictive Tool for Drug Design, Discovery, Delivery, and Disposition: Basic Concepts and Modeling

1. Artificial Neural Network

2. ANN as a Classifier

3. ANNs in Drug Delivery and Disposition

4. Applications of ANN Modeling in Drug Delivery and Pharmaceutical Research

5. ANN Applications in Analytical Data Analysis and Structure Retention Relationship Methodology

6. ANN Application in Preformulation and Optimization of Pharmaceutical Formulation

7. In Vitro In Vivo Correlation

8. ANN Applications in Predicting Blood–Brain Barrier Permeability

9. Quantitative Structure-Activity Relationships (QSAR) and Quantitative Structure-Property Relationships (QSPR)

Chapter 2. The Role of Artificial Neural Networks on Target Validation in Drug Discovery and Development

1. Introduction

2. Basics of ANN

3. Target Validation in Drug Discovery and Development

4. ANN in Target Validation in Drug Discovery and Development

5. Target Validation and Neural Networks

6. ANNs in ADME, Toxicity, and Drug Delivery

7. Summary

Chapter 3. Computational Basis of Neural Elements

1. Introduction

2. Overview of Artificial Neural Networks

3. Historical Perspectives

4. Characteristics of the Artificial Neural Network

5. Brain Networks

6. Challenges of Artificial Neuronal Networks

7. Neurons

8. Neuronal Processes

9. Synaptic Connectivity

10. Adaptation in the Brain

11. Microneural Network Themes

12. Lateral Inhibition

13. Sensory Transduction

14. Feedforward Excitation

15. Feedforward Inhibition

16. Feedback Excitation

17. Feedback Inhibition

18. Feedback/Recurrent Excitation

19. Feedback/Recurrent Inhibition

20. Neural Development

21. Genetic and Molecular Elements of Neural Development

22. Cortical Synaptogenesis

23. Cortical Organization

24. Cerebral Cortex

25. Sensory Cortex

26. Primary Sensory Cortex

27. Secondary Sensory Cortex

28. Motor Cortex

29. Association Cortex

30. Cerebral Dominance

31. Functional Basis of Cortical Development

32. Conclusions

Chapter 4. Genetic Algorithm Optimization of Bayesian-Regularized Artificial Neural Networks in Drug Design

1. Introduction

2. Genetic Algorithm Implementations in Drug Design QSAR

3. Bayesian-Regularized Artificial Genetic Neural Networks

4. Model's Validation

5. Datasets Sources and Preparation

6. BRGNN in Drug Design QSAR

7. Conclusions

Chapter 5. Neurobiological Computation and Neural Networks

1. Cognitive Neuroscience and New Technologies

2. Cells in the Nervous System and Information Processing

3. Neuroglias and Biological Synthesis of Information

4. Genetics and Cognition

5. Complexity of Information

6. Information Processing

7. The Brain and Complex Problem Resolutions

8. Emotions and Problem Solving

9. Intercerebral Connectivity and Emotions

10. Attention and Time in Information Processing

11. Importance of the Problem Resolution Process

12. Problem Solving and Movement

13. Relevance of Information

14. Memory and Problem Solving

15. Unconscious Reasoning and Complex Problem Solving

Section 2. Basics and Application of ANN in Drug Discovery

Chapter 6. Application of Artificial Neural Networks in Modern Drug Discovery

1. Introduction

2. Structure of an ANN

3. Information Processing in ANNs

4. Types of ANNs

5. Features of ANNs

6. Applications of ANNs in Drug Discovery

7. Conclusion

Chapter 7. Impact and Challenges of Chemoinformatics in Drug Discovery

1. Introduction

2. Conclusion

Chapter 8. Impact of Artificial Neural Networks in QSAR and Computational Modeling

1. Introduction

2. ANN-Based Modeling Studies

3. Concluding Remarks

Chapter 9. Data Mining in Drug Discovery and Design

1. Introduction

2. From High-Throughput Screening to Virtual Screening

3. Quantitative Structure–Activity Relationship

4. Pharmacokinetic and Pharmacodynamics

5. Conclusion

Chapter 10. Artificial Neural Networking in Controlled Drug Delivery

1. Introduction to Controlled Drug Delivery Systems

2. What are ANNs?

3. How to Use ANN and Other Related Techniques to Design CRDDS

4. Applications of ANNs in the Design of CRDDS

5. Limitations of ANN

6. Conclusion

Section 3. ANN in Drug Delivery

Chapter 11. Artificial Neural Networks in Drug Transport Modeling and Simulation–I

1. Introduction

2. Why Are ANNs, Modeling and Simulation of Drug Transport Useful?

3. ANNs and Their Helping Hand in the Discovery of New Drugs

4. Applicability to Research—Review of Scholarly Works

5. The Use of ANNs during Formulation Design

6. Conclusions

Chapter 12. Artificial Neural Networks in Drug Transport Modeling and Simulation–II

1. Introduction

2. Overview of Artificial Neural Networks

3. Modeling Smart Drug Transport Systems for Controlled-Release Technology

4. Simulation of STS and CRT Systems Using ANNs

5. Designing ANNs for Drug Transport Models

6. Conclusion

Chapter 13. Artificial Neural Network as Helping Tool for Drug Formulation and Drug Administration Strategies

1. Introduction and a Brief History

2. ANNs: A Brief Overview

3. Advantages of ANNs over Conventional Statistics

4. Development of an ANN Model

5. Application of ANNs in Pharmaceutical Formulation Development

6. Software Used for ANN Application

7. ANNs as a Helping Tool for Drug Administration Strategies

8. Conclusion

Chapter 14. ANN in Pharmaceutical Product and Process Development

1. Introduction

2. Background

3. ANNs in Pharmaceutical Product and Process Development

4. Conclusion

Section 4. ANN in Drug Disposition

Chapter 15. Classic Formal Logic and Nonclassical Logics: Basis of Research on Neural Networks

1. Logic and Neural Networks

2. Nonformal Logic and Reasoning

3. Classical Formal Logic

4. Formal Logic Principles

5. Propositional Logic and Reasoning

6. Binary Logic and Decision-Making

7. Predicate Logic and Reasoning

8. Polyadic Predicates and Relationship Systems

9. Notions of Relationships of Higher Order

10. Properties of Systems of Relations

11. Representation: Summary of Classical Formal Logic

12. Nonclassical Formal Logic: A New Reasoning Process

13. Extensive Logics of Classical Nonformal Logic

14. Disruptive Logics of Classical Formal Logic

15. Nonclassical Logics, Math, and Neural Networks

16. To Conclude

17. Representation: Summary of Nonclassical Logics

Chapter 16. Neural Networks and Computational Complexity

1. P–NP Problems

2. Heuristics and Problem-Solving

3. Metaheuristics and Complex Problems

4. Synthetic Representation of Metaheuristics

Chapter 17. Adaptive Modeling and Intelligent Control of a Sodium Nitroprusside Delivery System

1. Introduction

2. Generalized Fuzzy Neural Network

3. Adaptive Modeling of the SNP Delivery Systems

4. G-FNN-Based Control of the SNP Delivery Systems

5. Simulation Results

6. Conclusions

Section 5. ANN in Various Applications in Medicine

Chapter 18. Recent Advances of Biochemical Analysis: ANN as a Tool for Earlier Cancer Detection and Treatment

1. Introduction

2. ANNs in Cancer Diagnostics

3. Design and Use of ANNs in a Clinical Setting

4. Conclusion and Future Perspectives

Chapter 19. Role of an Artificial Neural Network Classifier in Nuclear Pleomorphic Feature Analysis of Histopathological Images of Breast Cancer

1. Introduction

2. Digital Pathology

3. Nuclear Pleomorphism

4. Artificial Neural Networks

5. ANN Classifier

6. Classifier Optimization

7. Results and Discussion

8. NeuroSolutions Classifier

9. Proposed Idea

10. Future Work

11. Conclusion

Chapter 20. Clinical Applications of Artificial Neural Networks in Pharmacokinetic Modeling

1. Introduction

2. What Are ANNs?

3. ANNs in Pharmacokinetic and Pharmacodynamic Modeling

4. Clinical Application of ANNs

5. Conclusion

Index

Copyright

Academic Press is an imprint of Elsevier

125 London Wall, London EC2Y 5AS, UK

525 B Street, Suite 1800, San Diego, CA 92101-4495, USA

225 Wyman Street, Waltham, MA 02451, USA

The Boulevard, Langford Lane, Kidlington, Oxford OX5 1GB, UK

Copyright © 2016 Elsevier Inc. All rights reserved.

No part of this publication may be reproduced or transmitted in any form or by any means, electronic or mechanical, including photocopying, recording, or any information storage and retrieval system, without permission in writing from the publisher. Details on how to seek permission, further information about the Publisher’s permissions policies and our arrangements with organizations such as the Copyright Clearance Center and the Copyright Licensing Agency, can be found at our website: www.elsevier.com/permissions.

This book and the individual contributions contained in it are protected under copyright by the Publisher (other than as may be noted herein).

Notices

Knowledge and best practice in this field are constantly changing. As new research and experience broaden our understanding, changes in research methods, professional practices, or medical treatment may become necessary.

Practitioners and researchers must always rely on their own experience and knowledge in evaluating and using any information, methods, compounds, or experiments described herein. In using such information or methods they should be mindful of their own safety and the safety of others, including parties for whom they have a professional responsibility.

To the fullest extent of the law, neither the Publisher nor the authors, contributors, or editors, assume any liability for any injury and/or damage to persons or property as a matter of products liability, negligence or otherwise, or from any use or operation of any methods, products, instructions, or ideas contained in the material herein.

ISBN: 978-0-12-801559-9

British Library Cataloguing-in-Publication Data

A catalogue record for this book is available from the British Library

Library of Congress Cataloging-in-Publication Data

A catalog record for this book is available from the Library of Congress

For information on all Academic Press publications visit our website at http://store.elsevier.com/

Typeset by TNQ Books and Journals

www.tnq.co.in

Printed and bound in the United States of America

Dedication

This work is dedicated to my big family, without their patience and incredible constant support it wouldn't be possible to deliver this book. They all deserve recognition for shaping my life and giving me time.

Munish Puri, MS, PhD

To the loving memories of my parents memories of his parents, Dr Keshav Baliram Hedgewar, who gave proper direction, my beloved wife Seema, who gave positive meaning, and my son Sarvadaman, who gave a golden lining to my life.

Yashwant Pathak, PhD

This work is dedicated to the loving memories of my father who passed away on April 22, 2013.

Vijay Kumar Sutariya, PhD

I dedicate this book to my family, friends, and teachers and everybody who made an impact in my life.

My special acknowledgment goes to my wife Kiran and children Shraddha and Krishna, without their encouragement this would not have been possible.

Parents are our first teachers, and I am extremely fortunate to have parents that taught me a lot in my life. I would like to dedicate this to them.

Srinivas Tipparaju, PhD

To the Ibero-American Science and Technology Education Consortium (ISTEC) for giving us the platform and opportunity to meet and start a journey to improve education throughout the Ibero-American region. To our wives and offspring that have supported us throughout this journey that we not only have passion for but also are fully committed to.

Luis Fernando Cruz, PhD

Wilfrido Moreno, PhD

Contributors

Snezana Agatonovic-Kustrin,     MARA University of Technology Selangor, Malaysia

Orhan E. Arslan,     Department of Pathology and Cell Biology, University of South Florida Morsani College of Medicine, Tampa, FL, USA

Pruthvi Raj Bejugam,     National Centre for Cell Science, NCCS Complex, Pune University Campus, Pune, India

Jonathan Bernick,     Independent Consultant, Omaha, NE

Marilyn Bui

Pathology and Cell Biology, University of South Florida, Tampa, FL, USA

Analytic Microscopy, Moffitt Cancer Center, Tampa, FL, USA

Jeffrey Burgess,     Department of Pharmaceutical Sciences, College of Pharmacy, University of South Florida, Tampa, FL, USA

Julio Caballero,     Centro de Bioinformatica y Simulacion Molecular, Universidad de Talca, Talca, Chile

Tapash Chakraborty,     Department of Pharmaceutical Sciences, Dibrugarh University, Dibrugarh, India

Sharmistha P. Chatterjee

Engineering Technology & Computer Science, Broward College North Campus, Hindu University of America, Lighthouse Point, FL, USA

Department of Computer & Electrical Engineering and Computer Science, Florida Atlantic University, Boca Raton, FL, USA

Harsh Chauhan,     School of Pharmacy and Health Professions, Creighton University, Omaha, NE

Luis Fernando Cruz Quiroga,     Complex Systems & Education Network for the Ibero-American Science and Technology Education Consortium (SCED-ISTEC)

Malay K. Das,     Department of Pharmaceutical Sciences, Dibrugarh University, Dibrugarh, India

Pranab Jyoti Das,     Department of Pharmaceutical Sciences, Dibrugarh University, Assam, India

Todd Daviau,     CoreRx, Inc., Clearwater, FL, USA

Meng Joo Er,     School of Electrical and Electronic Engineering, Nanyang Technological University, Singapore

Michael Fernandez,     Virtual Nanoscience Laboratory, CSIRO Materials Science & Engineering, Parkville, VIC, Australia

Anastasia Groshev,     University of South Florida, Morsani College of Medicine, Tampa, FL, USA

Manish K. Gupta,     School of Pharmacy, Lloyd Institute of Management and Technology, Greater Noida, Uttar Pradesh, India

Swati Gupta,     School of Pharmaceutical Sciences, Apeejay Stya University, Gurgaon, Haryana, India

Syeda Saba Kareem,     Pharmacy Department, St. Joseph's Hospital, Tampa, FL, USA

Mark Lloyd,     Analytic Microscopy, Moffitt Cancer Center, Tampa, FL, USA

Matthew MacPherson,     Department of Chemical Engineering, College of Engineering, University of South Florida, Tampa, FL, USA

Vineetha Mandlik,     National Centre for Cell Science, NCCS Complex, Pune University Campus, Pune, India

Vijay Masand,     Department of Chemistry, Vidya Bharti College, Amravati, Maharashtra

Bhaskar Mazumder,     Department of Pharmaceutical Sciences, Dibrugarh University, Assam, India

Brain McMillan,     CoreRx, Inc., Clearwater, FL, USA

Wilfrido Alejandro Moreno

Department of Electrical Engineering, University of South Florida, Tampa, FL, USA

R&D of Ibero-American Science and Technology Education Consortium (ISTEC)

David Morton,     School of Pharmacy and Applied Science, La Trobe Institute of Molecular Sciences, La Trobe University, Bendigo, VIC, Australia

Timothy Padawer,     Department of Pharmaceutical Sciences, College of Pharmacy, University of South Florida, Tampa, FL, USA

Abhijit S. Pandya,     Department of Computer & Electrical Engineering and Computer Science, Florida Atlantic University, Boca Raton, FL, USA

Jayvadan Patel,     Nootan Pharmacy College, S.K. Patel Campus, Visnagar, Gujarat, India

Anita Patel,     Nootan Pharmacy College, S.K. Patel Campus, Visnagar, Gujarat, India

Yashwant Pathak,     USF College of Pharmacy, University of South Florida, Tampa, FL, USA

Dev Prasad,     Formulation development, Fresenius Kabi USA, Skokie, IL

Charles Preuss,     College of Medicine, University of South Florida, FL, USA

Munish Puri

Electrical Engineering, University of South Florida, Tampa, FL, USA

Analytic Microscopy, Moffitt Cancer Center, Tampa, FL, USA

Visiting Fellow, National Cancer Institute, NIH, Bethesda, MD, USA

Ravindra K. Rawal,     Department of Pharmaceutical Chemistry, ISF College of Pharmacy, Moga, Punjab, India

Shailza Singh,     National Centre for Cell Science, NCCS Complex, Pune University Campus, Pune, India

Aum Solanki,     Department of Pharmaceutical Sciences, College of Pharmacy, University of South Florida, Tampa, FL, USA

Pochi R. Subbarayan,     Department of Medicine, Division of Hematology and Oncology, University of Miami Miller School of Medicine, Miami, FL, USA

Srinivas M. Tipparaju,     Department of Pharmaceutical Sciences, College of Pharmacy, University of South Florida, Tampa, FL, USA

Yong Zhang,     School of Electrical and Electronic Engineering, Nanyang Technological University, Singapore

Foreword

The drug discovery research and development process (R&D) is complex. A successful drug discovery requires the best scientific minds and expertise focusing on resource and task management under significant time and cost constraints. From a system design point of view, complexity arises from the considerable underlying uncertainties and the wide range of dependent and independent variables. Consider that the potential target for a new drug design must be discerned from a database of 20,000 to 25,000 human genes made up of three billion individual base pairs to match the right binding pocket of the target protein.

The drug discovery R&D process is a multiobjective parallel challenge at both the in silico and the experimental levels. The capabilities of available computational tools have been realized in the emerging areas of combinatorial chemistry and high-throughput screening to handle large data of over 35 million chemical compounds and their probable physical, chemical, and structural properties. Improvements in computational capabilities—such as those due to the application of artificial neural networks (ANNs)—are continuing to work their way into the state of the art and are positively contributing to the challenges of addressing complex, multiobjective optimization and focused searches in uncertain and complex environments. The application of ANNs to the drug discovery R&D process is the main topic of this book.

ANNs are computational learning machine networks—inspired by human brain neurons—that utilize nonlinear mapping techniques. As is evident from the chapters in this book, ANNs are very suitable for application to drug discovery R&D. They employ highly parallel processing techniques that can address complex system environments characterized by a high degree of uncertainty over a wide range of independent variables and many dependent variables and can be used as a predictive tool as they learn from past experiences and adapt. The ANN provides a pathway to tackle significant time and cost constraints due to its ability to address the nonlinear and nonparametric nature of the problems.

Computational model developers, researchers, medicinal chemists, and the many other experts that contribute to the highly sequential process of drug discovery and development will find this text particularly useful. As presented here in 19 chapters prepared by many of the world's leading experts, the ANN plays a central role in drug design, discovery, delivery, and disposition.

The 19 chapters are assembled into five main sections:

Section I: Basics of ANN: Concept and Strategy in Drug Design

Section II: Basics and Application of ANN in Drug Discovery

Section III: ANN in Drug Delivery

Section IV: ANN in Drug Disposition

Section V: ANN in Various Applications in Medicine

These chapters include, but are not limited to, coverage of basic concepts and modeling, the role of ANNs in target validation, genetic algorithm optimization in drug design, neurobiological computation, challenges in Chemoinformatics, the impact of ANN in quantitative structure–activity relationship and in computational modeling, data mining in drug discovery and design, drug transport modeling and simulation, drug formulation and drug administration strategies, pharmaceutical product and process development, computational complexity, adaptive modeling and intelligent control, and cancer detection and treatment.

I would like to congratulate the Editors for preparing this outstanding multidisciplinary text that bridges the space between engineering and health sciences. Dr. Moreno and Dr. Puri—representing engineering—and Dr. Sutariya, Dr. Tipparaju, and Dr. Pathak—representing the health sciences—have worked together to create what is sure to become a standard reference for drug discovery R&D researchers, students, and professionals.

I expect that the scientific community worldwide will welcome this effort!

Dr. Robert H. Bishop, P.E.,     Dean, College of Engineering The University of South Florida

Preface

Modern drug discovery is an outcome of collaborative and cooperative efforts at the level of researchers in academic, industry, and government research institutions. Computational processing and molecular modeling help scientists to harness their knowledge gained from recent advances in genomics and proteomics to understand biological systems and disorders as well as diseases.

The research and development (R&D) process is a complex and challenging task that involves resource and task management, the best scientific minds and expertise, time factors, and cost. The complexity of the biological system starts from the potential target for a new drug design from the database of 20,000 to 25,000 human genes made up of three billion individual base pairs to match a right binding pocket of the target protein. Target validation is a complex and crucial step in drug development that helps scientist to avoid any frustrating dead end research pathways. Medicinal chemists optimize the lead compound to become a potential drug by understanding the structural parameters of the target. The role of technology, computational tools, and smart algorithms is very crucial at these earlier stages of drug development. Any mistakes and wrong assessments in prioritizing the lead compound may affect cost, time, and research efforts.

Artificial neural networks (ANNs) are widely used for various biomedical applications like computational chemistry at the molecular level, bioinformatics, Chemoinformatics, and quantitative structure–activity relationships (QSARs). Books available on ANN express the theoretical or technical aspects of the mathematical modeling involved in the ANN approach to solve a problem in Chemoinformatics. The use of ANN in medicinal chemistry is quite common. Computer-aided designs of drugs use computational methods to design ligands and structure-based drug designs. Based on the binding pocket affinity of the target and database techniques, the optimization of design parameters are predicted in mathematical modeling. Computational methods are used to understand diseases at the molecular level and to design a safe and effective drug. Molecular mechanics are a helping tool to predict the conformal changes in target-based quantitative models and binding affinity in the whole design process. Computational models are often structured around the ligand–protein interaction and the target's structural parameters-based analysis.

It takes 8–15  years to develop a new drug from the time it is discovered (discovery time less than 1  year) to making it available in the market. The important factor other than time is the cost involved in the R&D of drug development (including experimental failures), estimated around $800m to $1b. The mounting cost of phase II and phase III trials and reducing attrition rates are additional challenges for the pharmaceutical industry. For every 5000 to 10,000 compounds in the R&D process, only one receives FDA approval.

ANN is widely used in QSAR in situations when the dataset is very large and cannot solve with linear functions. Multilayers are designed as hidden neuron layers with a varying set of neuron numbers. Input variables are selected from the information related to the drug’s parameters like concentration, compounds, ligands, etc. and processed through multiple training steps to meet the output. The predicted value is analyzed and compared with the known value. The difference in predicted and known values is propagated backward until the difference becomes negligibly small to achieve the estimated drug values.

Based on the above discussion we believe that there is a need for a reference book that will address various aspects of ANN and its applications in drug design, delivery, and dispositions. After carefully studying the literature we found that there are several books available on the market related to computational drug design, molecular modeling, Chemoinformatics on the drug design side, and ANN applications in biomedical, cancer, cardiovascular, and mathematical modeling in neuroscience. However, there is no consolidated reference book that discusses the applications of ANN in drug design, delivery, and disposition.

This interface of ANN from the computational engineering side and drug design, discovery, delivery, and disposition from the medicine side will not only solve the problem of an 8–15  years-long period of drug design and development but also will give a paradigm shift in designing new models and help in designing a predictive tool for an effective drug development and disposition system.

ANN's big advantage of learning and self-correcting ability even in a highly nonlinear, complex, noisy environment will be a milestone in the direction of drug designing and controlled delivery for the future pharmaceutical industry. The ANN can predict to deliver the drug even in deep brain areas with the help of implants in Parkinson's and epilepsy diseases. Our book will be a unique, knowledgeable resource for the researchers, scientists, and academics working in the medicine and computational modeling communities and will be a trendsetter in this field.

This book focuses on the following major aspects of ANN used in drug design, drug discovery, delivery, and disposition:

Section I: Basics of ANN: Concept and Strategy in Drug Design

Section II: Basics and Application of ANN in Drug Discovery

Section III: ANN in Drug Delivery

Section IV: ANN in Drug Disposition

Section V: ANN in Various Applications in Medicine

The chapter authors are leading scientists in their respective fields who have contributed significantly in this field.

The chapters focus on the ANN in drug design, delivery, adopted methodologies, modeling, and their applications in treatments of various diseases and disorders. The book will be an excellent resource for scientists and graduate students who are working in the field of ANNs, especially in the area of ANN applications in drug design, discovery, delivery, and disposition.

We believe that this book fills a gap and will be welcomed by the scientific community worldwide.

Munish Puri

Yashwant Pathak

Vijay Kumar Sutariya

Srinivas Tipparaju

Wilfrido Moreno

Section 1

Basics of ANN: Concept and Strategy in Drug Design

Outline

Chapter 1. Introduction to Artificial Neural Network (ANN) as a Predictive Tool for Drug Design, Discovery, Delivery, and Disposition: Basic Concepts and Modeling

Chapter 2. The Role of Artificial Neural Networks on Target Validation in Drug Discovery and Development

Chapter 3. Computational Basis of Neural Elements

Chapter 4. Genetic Algorithm Optimization of Bayesian-Regularized Artificial Neural Networks in Drug Design

Chapter 5. Neurobiological Computation and Neural Networks

Chapter 1

Introduction to Artificial Neural Network (ANN) as a Predictive Tool for Drug Design, Discovery, Delivery, and Disposition

Basic Concepts and Modeling

Munish Puri¹, Aum Solanki², Timothy Padawer², Srinivas M. Tipparaju², Wilfrido Alejandro Moreno³,  and Yashwant Pathak⁴     ¹Visiting Fellow, National Cancer Institute, NIH, Bethesda, MD, USA     ²Department of Pharmaceutical Sciences, College of Pharmacy, University of South Florida, Tampa, FL, USA     ³Department of Electrical Engineering, University of South Florida, Tampa, FL, USA; and R&D of Ibero-American Science and Technology Education Consortium (ISTEC)     ⁴USF College of Pharmacy, University of South Florida, Tampa, FL, USA

Abstract

An ANN is a computational model inspired by networks of biological neurons, wherein the neurons compute output values from inputs. It learns from its past experience and errors in a non-linear parallel processing manner. The neuron is the basic calculating entities which computes from a number of inputs and deliver one output comparing with threshold value and turned on (fired). The computational processing is done by internal structural arrangement consists of hidden layers which utilizes the back propagation and feed forward mechanism to deliver output close to accuracy. The learning is based on reinforcement (supervised) and unsupervised (no target) type. The unsupervised mimics the biological neuron pattern of learning. ANN is widely used in medicine as classifier in cancer applications, drug delivery, pharmaceutical research and Blood-Brain Barrier (BBB) permeability.

Keywords

Artificial Neural Network; ANN in medicine; ANN as classifier; neural network in pharmaceutical; ANN in Blood-Brain Barrier (BBB) permeability

Contents

1. Artificial Neural Network 3

2. ANN as a Classifier 5

3. ANNs in Drug Delivery and Disposition 6

4. Applications of ANN Modeling in Drug Delivery and Pharmaceutical Research 7

5. ANN Applications in Analytical Data Analysis and Structure Retention Relationship Methodology 7

6. ANN Application in Preformulation and Optimization of Pharmaceutical Formulation 8

7. In Vitro In Vivo Correlation 9

8. ANN Applications in Predicting Blood–Brain Barrier Permeability 10

9. Quantitative Structure-Activity Relationships (QSAR) and Quantitative Structure-Property Relationships (QSPR) 11

1. Artificial Neural Network

An Artificial Neural Network (ANN) is a computational model inspired by networks of biological neurons, wherein the neurons compute output values from inputs. All signals can be assigned binary values as either 1 or −1. The neuron calculates a weighted sum of inputs and compares it to a threshold of 0. If the calculated sum is higher than the threshold, the output is set to 1 or to −1. The power of the neuron results from its collective behavior in a network where all neurons are interconnected. The network starts evolving: neurons continuously evaluate their output by looking at their inputs, calculating the weighted sum, and then comparing to a threshold to decide if they should fire. This is a highly complex parallel process whose features cannot be reduced to phenomena taking place with individual neurons. One observation is that the evolution of an ANN causes it to eventually reach a state where all neurons continue working, but no further changes in their state occur. A network may have more than one stable state, and it is determined by the choice of synaptic weights and thresholds for the neurons.

ANN is a computational model that is based on a machine learning technique. It works like a human brain neuron system. This machine learning technique follows the same pattern of learning, that is, learning from its past experience and mistakes like mammalian neurons to achieve the target value. An algorithm is designed on the basis of a neural network system to implement a parallel computational power of neurons. ANN learns from its past experience and errors in a nonlinear parallel processing manner using a popular algorithm named feed forward and backpropagation. The term feed forward describes how the neural network processes and recalls patterns. In a feed forward neural network, neurons are only connected forward. Each layer of the neural network contains connections to the next layer, but there are no connections back. The term backpropagation describes how this type of neural network is trained. Backpropagation is a form of supervised training. When using a supervised training method, the network must be provided with both sample inputs and anticipated outputs. The anticipated outputs are compared against the actual outputs for given input. Using the anticipated outputs, the backpropagation training algorithm takes a calculated error and adjusts the weights of the various layers backward from the output layer to the input layer to reduce the value of error. The information is delivered to output if it achieves the target; otherwise, it is backpropagated. Hence the name of the algorithm is feed forward backpropagation. The target value will only be achieved if the weighted sum will meet the minimum threshold and hence feed forward or backpropagate for further processing. ANN could be an excellent choice to process large biological data for a more accurate prognosis. The prognostic tools can be designed based on ANN's powerful learning and processing characteristics, which can work perfectly even in a highly probabilistic and noisy environment. The power of the neuron results from its collective behavior in a network where all neurons are interconnected. The network starts evolving; neurons continuously evaluate their output by looking at their inputs, calculating the weighted sum, and then comparing to a threshold to decide if they should fire.

The neuron is the basic calculating entity in ANN processing, which accepts information from a number of inputs and delivers one output by comparing with a threshold value. The computational processing is accomplished by internal structural arrangements that consist of hidden layers and algorithms to deliver a specified output. The learning is based on reinforcement (supervised) and unsupervised (no target) types. The unsupervised mimics the biological neuron pattern of learning.

Basically, ANN is a mathematical model that is used to implement the designed algorithm-based machine learning techniques. ANN communication is performed by calculating the weights of neural inputs, which works on the basis of mathematical operations such as multiplication and addition. Each input received at a nodal point is multiplied with its weights and summed together before activation (firing). In the case of a biological neuron, the information is received at dendrites, processed at soma (cell body), and delivered to axon (output). Similarly, in ANN, the artificial neuron is the basic unit of information reception where the inputs are received and multiplied, summed, and processed via a transfer function before being delivered to the output. An ANN model is so simple and natural that it can handle very complex real-life problems in a nonparallel and distributive way like a biological neural network. The mathematical description of ANN can be understood by the following equation:

(1)

where

Xi(t) is the input value at time t,

Wi(t) is the weight of neural input at time t,

c is the bias,

F is a transfer function,

Y(t) is the output value at time t.

Note that the transfer function F is to be selected on the basis of the nature of the problem. It mathematically defines the properties of neurons. It can be any step function or nonlinear sigmoid function, depending on the problem. The step function is used to handle classification problems like classifying the benign and malignant state of breast tumors. Similar to a human neuron network, ANN should be trained before it is actually applied to a specific problem. This learning can be supervised or unsupervised in nature.

2. ANN as a Classifier

ANN can be used to classify the complex and noisy biological data for prognosis. For example, in breast cancer tumor data, an ANN classifier can be trained to classify the benign and malignancy states based on descriptors like cell uniformity, clump thickness, size, shape, intensity, mitosis, etc. Its performance is then judged through mean square error and confusion matrices. Data are loaded in terms of feature elements.

The data available for processing under ANN are distributed in categories such as training, validation, and testing. Training data are the actual data offered to a network during training and adjusted according to errors and mistakes. Validation data are used to test the network performance directly and stop the processing in case of overfitting. Out of sample testing is an independent operation and has no effect on ANN operation during training. The hidden neuron number can be adjusted at any level if the performance is unsatisfactory.

For example, in the case of breast cancer histopathology image analysis, the process of classifier design and testing starts with data collection. The high-resolution cancer images are processed for segmentation using various mathematical models and algorithms. Segmentation techniques like adaptive thresholding, water shedding, nearest neighbor, etc. are mainly used before feeding data to a specific classifier. Pleomorphic (distorted shapes and sizes) nuclei are then processed for feature extraction like size, shape, textures, etc. under a selective morphometric parameter check. Extracted features are classified under the neural network algorithm. For next generation healthcare solutions, ANN computational processing algorithms could be a powerful tool to assist pathologists in second opinions, efficient drug design, and delivery in the pharmacy industry in an environment of rapid decision making, reducing the drug discovery time from lab to market.

3. ANNs in Drug Delivery and Disposition

The focus of this chapter is to discuss how ANNs can be applied to drug delivery and disposition. ANNs provide a virtual method to predict the effects the drug will have on the body, with the ultimate goal of improving patient care by streamlining the development of pharmaceuticals. The development of ANNs in drug disposition is an interdisciplinary and complex area, involving pharmacologists and computational biologists. The power of ANNs is that they can minimize the need for traditional pharmacology, including animal studies, therefore reducing time and costs for drug discovery and developmental research. Software can be used to not only develop new drugs but also improve current products. However, the precision of ANNs in drug disposition is still under development, and newer research and tools are evolving. It is hoped that this area of research will grow along with high throughput technologies capable of generating omics data. The use of biological data gained in this postgenomic era has opened up a new field referred to as systems pharmacology. A large variety of biological databases exist and offer researchers access to a wide range of information, including genomic, proteomic, metabolomics, drug–protein interactions, and gene expression. It is possible that these datasets can be utilized as input for ANNs for answering various questions in pharmacology [1].

Pharmaceutical therapies take many forms, that is, tablets, liquids, and creams. These compounds are intended to reach a specific target in order to have the desired treatment effect. In designing ANNs for drug disposition, the physiochemical properties of the active pharmaceutical ingredients (API) and excipients are of primary importance, along with the processes they undergo in order to reach their targets [2]. For example, the efficacy of an orally administered drug will depend on its ability to be efficiently absorbed by the gastrointestinal (GI) tract. Traditional in vitro dissolution tests face difficulties in accurately predicting the physiological conditions of the upper GI. The conditions of the GI are also dynamic and can vary greatly between fasted and fed states. The differences in these conditions can have a wide effect on drug disposition [3]. ANNs for use in drug transport simulation are constructed to best represent the anatomical system at each step of the absorption process. This type of modeling simplifies the complex nature in which compounds interact with the body, starting at the target site of ingestion or application to the therapeutic site. The ability for the drug to reach specific compartments of the body must be considered when modeling drug delivery. There are a number of software packages available that simulate drug transport in the gastrointestinal tract, such as Intellipharm™, GastroPlus™, and PK-Sim®. The section on drug disposition consists of chapters dealing with this area and further discusses the state of the art in the field with examples of various therapies. Specific examples include challenges faced with advanced drug delivery and disposition to specialized targets, such as ocular drug delivery and transportation, to solid oral tablet drug transport for hypercholesterolemia treatment. It gives insights into how ANNs can be used to model each type of therapy.

4. Applications of ANN Modeling in Drug Delivery and Pharmaceutical Research

The modeling and nonlinear pattern recognition abilities of ANNs increase their potential applications in many different pharmaceutical research studies. These applications include drug modeling, pharmacokinetic and pharmacodynamics modeling, interpretation of analytical data, and, especially, in vitro/in vivo correlations. Also, ANNs have been used in the field of biopharmaceutical production. Takahashi et al. developed an ANN model to predict molecular and viable cell concentrations based on previously published UV–visible spectral data [4]. They were able to develop a suitable 4-layered model that can allow researchers to use UV–visible spectroscopy in order to characterize cell culture broths instead of using more expensive experimental techniques [4].

5. ANN Applications in Analytical Data Analysis and Structure Retention Relationship Methodology

ANNs can be very useful in predicting the relationship between the structural properties of a molecule and its behavior in biological and chromatographic environments. These Structure Retention Relationship (SRRs) can be used to predict retention for a new solute, identify the most significant structural descriptors of molecules, gain insight into the molecular mechanism of chromatographic separation, evaluate physiochemical properties of molecules, and predict biological activities within a set of drugs and other xenobiotics [5]. Tham et al. used ANNs to predict the retention capabilities of 18 amino acids in reversed-phase, high-performance liquid chromatography (HPLC) [6]. After encoding the amino acids with 36 molecular descriptors, the genetic neural network method was used to establish the correlation between the molecular descriptors and retention time. Through a genetic algorithm, it was determined that a 5-2-1 architecture was the best suited for predicting retention, which revealed that there is a strong link between the structure of amino acids and chromatographic separation [6].

D'Archivio et al. used ANNs' applications in SRR studies to create a model predicting the Kováts retention indices (RI) of 90 different esters, separated into seven different groups (columns) based on their polarity [7]. Here again, ANNs showed their proficiency in nonlinear prediction. The performance of the ANN model was compared to that of a multiple linear regression (MLR) model. Both models predicted cross-column RI values comparably. However, when less similar columns were used, the nonlinear ANN model held a definite advantage, showing its ability to be applied to a wider range of data [7].

Such applications for SRR studies can be conducted on different natural products as well. HPLC, working in combination with high-resolution mass spectroscopy, is generally used to separate natural extracts. Eugster et al. developed an ANN model using a chemical dataset of 260 different natural products [8]. They found that an 8-5-1 architecture was the best-suited model in terms of predictive ability [8]. This predictive ability of the ANN model was compared to that of partial least square (PLS) regressions. PLS regressions required the use of multiple models that operated only on certain subsets of the 260 natural products, whereas the ANN model was able to operate with comparable predictive ability on the entire set of natural products, again showing the versatility of ANN models [8].

6. ANN Application in Preformulation and Optimization of Pharmaceutical Formulation

ANNs can be a useful tool for preformulation studies, as they can be used to predict the chemical properties of different compounds. Ebube et al. used ANNs to predict physiochemical properties of amorphous polymers [9]. The model was able to accurately predict (error of 0–8%) the relationships between chemical composition of the polymer and the water uptake profiles, viscosity, and glass transition temperatures [9]. These results show the potential that the ANN model has as a preformulation tool.

The nonlinear pattern recognition abilities of ANN models can also be incredibly useful in the optimization of pharmaceutical formulations and dosage methods. ANNs have been shown to be capable predictors of formulations' in vivo dissolution time profiles and bioavailability profiles, making it possible to identify formulations with such desired characteristics [10]. Subramanian et al. compared the optimization ability of ANNs to that of MLR analyses [11]. An ANN model and 3³ factorial design model were used to optimize the formulation parameters of cytarabine liposomes. A set of input variables was used with 11 hidden layers and one output variable (percentage drug entrapment). Optimal drug entrapment was found to occur in a 1:13 drug-to-lipid ratio. The optimization was then validated by preparing additional formulations, revealing that the ANN model provides more accurate predictions [11]. Kumar et al. used ANNs to optimize fatty alcohol concentration in oil/water emulsions [12]. An R² value of 0.84 shows that the ANN model was successful in predicting the optimal concentration. In order to optimize the polydispersity index of acetaminophen nanosuspensions, Aghajani et al. used an ANN model with independent variables: surfactant concentration, solvent temperature, and flow rate of solvent and antisolvent [13]. Using the model, it was determined that low polydispersity index (PDI) can be obtained with high antisolvent flow rates and solvent temperatures and low solvent flow rates. ANNs have also shown potential as optimization tools for time-dependent pharmaceutical formulations [14].

7. In Vitro In Vivo Correlation

The most promising use of the ANN model is, perhaps, its ability to predict in vivo correlations from in vitro results. Here again, it is the nonlinear modeling ability of the ANN model that separates it from other regression methods. Dowell et al. studied the ability of different ANN models as in vitro in vivo correlation (IVIVC) tools. For simpler sets of data, the feed forward neural network (FFNN) and the generalized regression neural network (GRNN) were the best because of their broad predictive ability, whereas for more complex datasets, it may be necessary to use other types of neural networks such as jump connection neural networks or recurrent neural networks [15]. Regardless, ANNs show a great applicability for IVIVCs because they avoid the need to provide a priori regression equations while having great predictive ability for complex, multivariable relationships [15]. Parojcic et al. compared the ability of the GRNN model to that of deconvolution and convolution approaches in predicting IVIVC for the drug release of paracetamol matrix tablets [16]. Findings showed that the GRNN model was superior to both approaches, resulting in a better IVIVC in a model that is easier to use for further research [16].

IVIVC for orally inhaled drugs is difficult to obtain, as there are more variables affecting the drug's efficacy, such as lung deposition and total systemic bioavailability, making ANNs great candidates as possible solutions [17]. De Matas et al. developed an ANN model that predicted the in vivo efficacy of dry powder inhaler formulations through the use of in vitro data and parameters based on the physiological characteristics of human volunteers [17]. Results showed that the ANN model is a very effective IVIVC tool for inhaled drugs (R²  ≈  80%), even though many improvements to the model can be made through the addition of significant input variables, larger datasets, and greater numbers of subjects [17]. A similar study developed an IVIVC for mild/moderate asthmatics receiving monodisperse salbutamol sulfate aerosols, with seven input variables and one hidden layer [18]. Here again, the ANN model showed its ability to be an effective predictive tool with results revealing the significance that aerodynamic particle size has on patient bronchodilator responses [18]. ANN models have also shown the ability to develop IVIVC for self-emulsifying delivery systems, dissolution kinetics in the GI tract, and metabolic clearance for new drugs [19–21].

The reason IVIVCs are so desirable for pharmaceutical research is that they provide preliminary information regarding the drug's in vivo behavior without conducting extensive in vivo experiments. Mendyk et al. developed a general IVIVC model that can be used as a preliminary predictor for the in vivo behavior for various pharmaceutical formulations through the use of detailed knowledge about their chemical and physiological characteristics [22]. This model has the potential to be used as a preliminary tool for researchers to study the bioavailability of new drug formulations.

8. ANN Applications in Predicting Blood–Brain Barrier Permeability

In order to be therapeutically effective, many drugs targeted