In silico methods and tools for drug discovery

Highlights

• In silico drug discovery methods are able to reduce the time and cost for drug discovery processes.

• Advancements in computational methods have enabled in silico drug discovery for practical use.

• Several FDA-approved drugs were developed with in silico methods.

Abstract

In the past, conventional drug discovery strategies have been successfully employed to develop new drugs, but the process from lead identification to clinical trials takes more than 12 years and costs approximately $1.8 billion USD on average. Recently, in silico approaches have been attracting considerable interest because of their potential to accelerate drug discovery in terms of time, labor, and costs. Many new drug compounds have been successfully developed using computational methods. In this review, we briefly introduce computational drug discovery strategies and outline up-to-date tools to perform the strategies as well as available knowledge bases for those who develop their own computational models. Finally, we introduce successful examples of anti-bacterial, anti-viral, and anti-cancer drug discoveries that were made using computational methods.

Introduction

Conventional drug discovery and development are risky, time-consuming processes that include target identification and validation, lead compound discovery and optimization, and preclinical and clinical trials [1]. In recent years, the estimated cost of bringing a new drug to market has reached about $1.8 billion USD [2], and the attrition rate of drug candidates is as high as 96% [2]. The reasons underlying this high attrition rate are poor drug efficacy and deficient drug absorption, distribution, metabolism, and excretion, and toxicity (ADME-Tox) [3]. Typically, in vivo and in vitro techniques are employed to examine drug safety, including adverse effects and toxicity. Recent advancements in in vitro models, such as organ-on-chip technology, have accelerated ADME-Tox assessments [4]. However, these approaches remain time-consuming, labor-intensive, and costly. High-throughput screening (HTS) methods have been developed to accelerate the identification of pharmacologically active chemical compounds from large numbers of molecules using automated assays [5]. Although automatic HTS systems reduce the need for human intervention, the scale of HTS remains low compared to the diversity of chemical structures. In addition, automated instruments remain expensive.

Recently, computer-aided drug discovery (CADD) approaches are attracting increasing attention as they can help mitigate the scale, time, and cost issues faced by conventional experimental approaches. CADD includes computational identification of potential drug targets, virtual screening of large chemical libraries for effective drug candidates, further optimization of candidate compounds, and in silico assessment of their potential toxicity. After these processes are conducted computationally, candidate compounds are subjected to in vitro/in vivo experiments for confirmation. Thus, CADD approaches can reduce the number of chemical compounds that must be evaluated experimentally while increasing the success rate by removing inefficient and toxic chemical compounds from consideration [6]. To date, CADD has been successfully employed to bring new drug compounds to market for diverse diseases, including human immunodeficiency virus (HIV)-1-inhibiting drugs (atazanavir [7], saquinavir [8], indinavir [9], and ritonavir [10]), anti-cancer drugs (raltitrexed [11]), and antibiotics (norfloxacin [12]).

Several CADD approaches have been developed and integrated with machine learning techniques to improve the accuracy and efficiency of CADD methods [13]. Structure-based drug discovery (SBDD) [14] and ligand-based drug discovery (LBDD) [15] are two different approaches taken in CADD. The selection of a suitable CADD approach relies on the availability of target protein structural information. To use the SBDD approach, structural information on the target protein is required, which is usually obtained experimentally by nuclear magnetic resonance or X-ray crystallography [14]. When neither is available, in silico prediction methods such as homology modeling [16] or ab initio modeling [17] can be used to predict the 3D structure of the target protein. Once the structure is available, structure-based virtual screening and molecular docking are possible [18]. When the structure is not available and it is not possible to predict a high-quality structure using in silico methods, the LBDD approach is often taken as an alternative. Although this approach requires prior information on the known active molecules of the target protein, many compounds have been discovered to treat diseases and are compiled in public databases unless the target is novel [[19], [20], [21]]. These approaches are introduced in section 4.

The field of CADD is rapidly advancing, and techniques and methods are under active development. Over the past few years, the integration of biological big data and machine learning approaches has opened new possibilities to increase the accuracy and efficiency of in silico drug discovery. This review introduces the overall procedures and methodologies behind in silico drug discovery, including target protein identification, chemical library screening, and toxicity assessment using machine learning approaches, summarizes available prediction tools and databases, and lists Federal Drug Administration (FDA)-approved and reported drug compounds developed using CADD techniques.