1
Deep Learning and Site-Specific Drug Delivery: The Future and Intelligent Decision Support for Pharmaceutical Manufacturing Science

Dhanalekshmi Unnikrishnan Meenakshi1*, Selvasudha Nandakumar2, Arul Prakash Francis3, Pushpa Sweety4, Shivkanya Fuloria5, Neeraj Kumar Fuloria5, Vetriselvan Subramaniyan6 and Shah Alam Khan1†

1College of Pharmacy, National University of Science and Technology, Muscat, Oman

2Department of Biotechnology, Pondicherry University, Puducherry, India

3Department of Biochemistry and Molecular Biology, Pondicherry University, Puducherry, India

4Anna University, BIT Campus, Tiruchirappalli, India

5Faculty of Pharmacy, AIMST University, Bedong, Malaysia

6Faculty of Medicine, Bioscience and Nursing, MAHSA University, Selangor, Malaysia

Abstract

Site-specific drug delivery [SSDD] is a smart localized and targeted delivery system that is used to improve drug efficiency, decrease drug-related toxicity, and prolong the duration of action by having protected interaction between a drug and the diseased tissue. SSDD system in association with the computational approaches is employed in discovery, design, and development of drugs to improve treatment outcomes. Artificial intelligence [AI] networks and tools are playing a prominent role in developing pharmaceutical products by employing fundamental paradigms. Among many computational techniques, deep learning [DL] technology utilizes artificial neural networks [ANN], belongs to machine learning [ML] approach that holds the key to measuring and forecasting a drug’s affinity for specific targets. It can reduce both cost and time by speeding up the drug development process rationally with careful decisions. DL is considered as the primary strategy to predict bioactivity as it shows improved performance compared with other technologies in the field. DL can assist in evaluating the success of a target-based drug design and development before the actual laboratory synthesis or production of the drug molecule. This chapter highlights the potential applications of DL in assigning a specific drug target site by predicting the structure of the target protein and drug affinity for a successful treatment. It also spotlights the impactful applications of many types of DL in SSDD and its advantages over conventional SSDD systems. Furthermore, some formulations that are intended to lead to the target or site-specific delivery and DL role in docking and pharmacokinetics profiling are also addressed. Ongoing challenges, skepticism about the likelihood of success, and the paths to overcome by future technological advancements are also dealt with briefly. Due emphasis is given to the use of DL in reducing the economic burden of pharmaceutical industries to overcome costly failures and in developing target specific new drug candidate[s] for a successful therapeutic regimen beneficial to human life.

Keywords: Site-specific, target, drug delivery, deep learning, machine learning, artificial intelligence, computational approach, precision medicine

1.1 Introduction

Site-specific drug delivery (SSDD) is an almost a century-old strategy but successful delivery of drugs to the target site without producing off-site unwanted adverse effects has not been realized yet. Random testing assays in the traditional development of SSDD identify only 3% of compounds that warrant further laboratory tests, and hence, it is vital to explore the drug-target interactions for every single pharmaceutical molecule. Modern drug discovery, which includes identifying and preparing drug-molecular targets with precision, is emerging to fill traditional SSDD gaps. Target-specific drug delivery promotes the delivery of medications to target sites without creating unwanted side effects elsewhere. Despite numerous publications and attention paid to the site-specific delivery that promises to “deliver” the medicine at the diseased site, the generation of target-specific therapeutic products has still been a challenge for researchers [1]. The obstructions met during the drug formulation process are mainly associated with the inability to foresee the impact of the combination of active pharmaceutical ingredients [APIs] and materials on the formulation parameters. A new drug formulation development process and the associated procedures need to satisfy the site-specific delivery and release profile. Moreover, it is a laborious task and the protocols to perform in vitro characterizations or modifications to obtain the desired profile are difficult for the formulators [2]. To bridge the knowledge gap and reduce the time required for selecting the best molecule for drug development, researchers have devised computational modeling approaches like molecular dynamics simulations, docking studies [3], and cheminformatics [4]. These helps in the evaluation of novel insights about the complex drug delivery systems, especially in atomic/cellular scale which experimental techniques cannot provide [5–7]. A revolution in data science has been observed in the last decades due to the usage of the graphics processing unit [GPU]. A large volume of drug-related data and techniques were generated and analyzed using artificial intelligence [AI] to predict drug interaction with the diseased targets in drug discovery. AI networks and tools are playing a prominent role in the development of pharmaceutical products by employing fundamental paradigms. In medicinal chemistry, several computational methods contributed to designing new drug candidates by relating the drug candidate’s physicochemical properties, biological activity, and binding affinity [8]. Machine learning [ML], the branch of AI, has gained importance in drug discovery protocols and has become the most attractive and prominent research areas. ML supports the advancement of effective formulation through data-driven predictions using experimental data. A well-designed ML technique can significantly speed up the optimization of formulations with reduced cost [9]. Knowledge acquisition about the molecular characteristics of lead molecules has been made with the help of ML techniques like partial least squares [PLS], k-nearest neighbors [kNN], and artificial neural networks [ANN] [10]. ANN is the most prevalent ML technique in formulation prediction [9, 10].

Among the various methodologies of AI, deep learning [DL] had gained significant attention in several areas because of its ability to extract features from data [11]. Leading pharmaceutical industries in collaboration with different AI organizations are trying to develop effective and ideal drug candidates in the field of oncology and CNS complications. In recent years, several trials involving the combination of nanotechnology and DL are underway to study their potential role in drug formulation with SSDD. The role of DL in drug development and manufacturing is depicted in Figure 1.1. DL methods are representation-learning techniques that can discover multiple-level representations of increasing complexity from the raw data using nonlinear models [12]. Several recent trials have connected nanotechnology and DL to study their potential role in drug formulation with site-specific drug delivery [SSDD]. DL can predict the probable drug carrier candidate through target-based drug designing and development. DL methods play a significant role in drug delivery by predicting (i) drug loading in the carrier, (ii) the enhancement in permeability through the body barriers, and choosing the stable drug delivery systems from different carriers and matrices [13].

Figure 1.1 Role of DL in drug development and manufacturing.

DL has proved to be an effective tool for virtual screening and predicting quantitative structure-activity relationships from large chemical libraries [14]. Golkov et al. reported that the DL is very useful in predicting the biological functions of several chemical compounds from the raw data based on their electronic arrangements [15]. A previous study on DL revealed that it has collected evidence from the vast amount of data sets related to the genome and utilized for drug repurposing and precise treatments [16]. Various DL models have been used to forecast interactions between protein-ligand, scoring docking poses, and virtual screenings. Thus, DL has been utilized to discover several endpoints in medicinal chemistry [17].

A study on predicting protein-ligand interactions using molecular fingerprints and protein sequences as vector input showed that the essential amino acid residues responsible for drug-target interactions were predicted using vectors obtained from the model [18]. A previous study by Lee et al. detailed a predictive model to represent the DeepConv-drug-target interactions [DTIs] in ligand-target complex. The predictive models were built using over 32,000 drug-target structures from the DrugBank, IUPHAR, and KEGG data sets. DNN outperforms similarity-based models and traditional protein representations, according to the findings [19]. For the prediction of novel DTIs between marketed medications and targets, Wen et al. used a successful DL method called deep belief network [DBN] and developed a methodology called DeepDTIs. This method was tested using an appropriate method and associated to suitable algorithms, such as random forest [RF], Bernoulli Naive Bayesian [BNB], and decision tree [DT]. Results showed that the algorithm used in this method...