The Role of Artificial Intelligence and Machine Learning in Modern Drug Discovery

According to the Precision Medicine Initiative, precision medicine is ‘‘an emerging approach for disease treatment and prevention that takes into account individual variability in genes, environment and lifestyle for each person” [1]. This new approach allows physicians and researchers to increase accuracy in predicting disease treatment and prevention strategies that will work for particular groups of people. This approach contrasts with the ‘‘one size-fits-all” approach, more widely used until relatively recently, in which the strategies mentioned above are developed with the average person in mind, regardless of differences between individuals.

The  opportunity for the creation of new treatments offeredby precision medicine generates at the same time great difficulties in the development of new methodologies. For this reason, in recent years a large amount of biomedical data has been generated, coming from very diverse sources: from small individual laboratories to large international initiatives. These data, known mostly as omic data (genomic, proteomic, metabolomic, pharmacogenomic, etc.), are an inexhaustible source of information for the scientific community, which allows stratifying patients, obtaining specific diagnoses or generating new treatments.

Diagnostic tests are frequently performed in some disease areas, as they allow immediate identification of the most effective treatment for a specific patient through a specific molecular analysis. With this, the practice of trial-and-error medicine, which is often frustrating and considerably more expensive, is often avoided. In addition, drugs created from these molecular characteristics usually improve treatment results and reduce side effects. One of the most common examples can be found in the treatment of patients with breast cancer. A significant percentage of patients with this type of tumor are characterized by overexpression of human epidermal growth factor receptor 2 (HER2). For these patients, treatment with the drug trastuzumab (Herceptin) in addition to chemotherapy treatment can reduce the risk of recurrence to more than 50%.

On the other hand, there are also the so-called pharmacoge nomic tests that provide assistance in making decisions related to the drug and the dose formulated for each patient. These decisions are based on the genomic profiles of the patients, so that they can metabolize certain drugs in different ways according to their genetics, thus causing adverse reactions. These reactions are related to variants in the genes that encode drug metabolizing enzymes, such as cytochrome P450 (CYP450). Pharmacogenomic testing can contribute to the safe and effective application of drugs in many different areas of health, including heart disease, cancer adjunctive therapy, psychiatry, HIV and other infectious diseases, dermatology, etc.

Table 1. An overview of additional studies that used machine learning for drug

Artificial intelligence in drug discovery

Enhanced computational power and the development of innovative techniques in the field of AI could be used to reform drug discovery and development processes. At the time of this literature review, the pharmaceutical industry is facing drops in efficiency of their drug improvement programs and simultaneous rises in research and development costs [23]. In recent years, there has been a radical expansion in the digitalization of information in the pharmaceutical industry; efficiently obtaining, examining, and applying this information to tackle complex clinical issues is a current challenge. AI can deal with enormous volumes of information with upgraded computerization. It can also integrate and use machine learning algorithms to increase efficiency and productivity. In this section, the main uses of AI to improve the effectiveness of the drug discovery cycle are discussed. above shows how and where AI can be integrated into drug discovery and development processes and the novel applications of AI in the pharmaceutical industry [25]. Drug discovery can be usefully split into four parts: drug design, polypharmacology, drug repurposing,  and drug screening. The primary use of AI is in predicting drug properties, which may reduce the need for clinical trials and live study participants, which would be beneficial from both financial and ethical standpoints. The studies identified in this review that support the integration of AI into the drug discovery procedure to improve efficiency, accuracy and productivity are discussed in this section.

Challenges  

1. Despite advances in AI and machine learning algorithm technologies implemented in the pharmaceutical industry, there are still many challenges regarding the implementation and integration of these technologies into the drug discovery process specifically and the pharmaceutical industry in general.
2. One problems is inefficient data integration. This problem results from diversity which exists between datasets, which may comprise raw data, processed data, metadata, or candidate data. These datasets must be collected and collated for efficient analysis, but currently, there is no established method of doing so. This is required before the drug discovery process begins, as without appropriately formatted data, the output of the machine learning algorithms will be inaccurate. More efficient methods for integrating available data into data banks before the drug discovery process begins are therefore required [38].
3. Another problem is occupational and skillset immobility: many people currently working in the pharmaceutical industry do not have the necessary skills or the qualifications needed to operate AI systems. Many people are proficient in data science, and others in molecular chemistry and biology, but few are experts in both, with the right combination of skills to apply AI in a pharmaceutical context. A knowledge of the underlying chemistry is required to generate appropriate algorithms, and vice versa [38].
4. A third but related difficulty is scepticism about machine learning and AI in the pharmaceutical industry owing to a lack of understanding on the methodology of algorithms, known as the “black box” phenomenon, and a lack of trust for the results generated. Those who are sceptical may be reluctant to use the data generated using AI and machine learning, wasting both time and money, and holding the industry back with regards to efficiency [24].
5. This distrust of AI leads to another problem: lack of financing for the development of AI in the pharmaceutical industry. Scepticism about the role and results of AI and machine learning in drug development processes may lead to hesitancy to invest money in this technology. This may lead to slower, less efficient research and development compared with its potential, in turn leading to a reduction in AI-related advances in the pharmaceutical industry. These are the different barriers which stand in the way of true development and these are the challenges which have to be overcome for AI to be integrated into the drug development processes. 

Artificial Intelligence or machine learning in precision drug discovery:

National Institute of health NIH has highlighted the fact that precision medicine is an emerging strategy for the ‘purpose of drug prevention or treatment which also considers the other variations in genetics, libertines and environment. This allows the doctors and physicians to treat and recover diseases more accurately than that of another method employed so far [25]. In order to make this more powerful it requires super mount and fastidious techniques which can be later used in an unfrequented way for trained set of data. The field of artificial learning uses the cognitive ability of the physicians and doctors using biomedical and bioinformatics data to produce fruitful results. Artificial intelligence can be broadly classified into almost three categories which include artificial general intelligence, artificial narrow intelligence and super intelligence [26]. ANI or artificial natural intelligence is still in the process of development and aims to hit the market or research in next decade. It has the ability to develop new data set, analyze them, to find the correlation among them and draw the meaningful or useful conclusions 

Future scope

The main potential of AI in the pharmaceutical industry is to reduce costs and increase efficiency [39]. Extensive research has demonstrated that dynamic learning can distinguish profoundly exact AI models while using half or less information than traditional AI and information subsampling approaches. Although the reason for this increased productivity is not fully understood, it appears that reduced repetition and predisposition, as well as gaining more significant information to traverse choice limits, are key components in this improved execution. As a result, without taking into account the expected mechanical overhead for actually carrying out dynamic learning efforts, screening expenses appear to be reduced by up to 90% [12].

Machine learning techniques can manage complex analyzes with huge, heterogeneous, and high-dimensional information collections with no manual input, which has proved helpful in the writing business applications. Combining machine learning, particularly deep learning, with human skill and experience might be the best way to coordinate numerous enormous data stores. The amazing information-mining capacity of AI innovation has given new essentiality to computer supported medication plans that incorporate multiple clinical considerations are better than piecemeal information, which can speed up prescription processes. With the further gathering of clinical data and the improvement of AI calculations, AI innovation is expected to enable many aspects of drug discovery and development, and become the standard computer supported medication plan strategy. The coordinated development of mechanization and innovations resulting from combining technologies should lead to advancements in medication resulting from improved analysis of large and complex datasets. This will be necessary to shorten drug development cycles, reduce costs, and improve success rates: the ultimate goal of implementing AI in this context .

AI in Pharmaceutical Analysis 

Pharmaceutical analysis encompasses the identification, determination, quantification, and purification of pharmaceutical substances, playing a crucial role in the drug discovery process. It primarily relies on qualitative and quantitative experimental methods. While these methods are highly accurate, the cost associated with screening new drug candidates from a vast array of natural products remains substantial. In contrast, computational methods offer a costeffective alternative. Therefore, AI technologies have been increasingly applied to enhance pharmaceutical analysis alongside traditional experimental methods.

Drug Toxicity Prediction

AI has become instrumental in advancing drug toxicity prediction, offering significant improvements in identifying potential adverse effects of novel drug candidates. Through meticulous training and validation, AI models adeptly delineate toxicity profiles, focusing on potential harm to specific organs or biological pathways. This capability enables the prioritization of compounds with minimized adverse effects, refining the selection of safer drug candidates 

Toxicity, a key indicator of a substance‘s harmful effects remains a central concern in drug development.[61] Traditional toxicity assessments often rely on in vivo animal testing, which not only raises ethical concerns but also significantly increases the cost of drug discovery.[61,119] In contrast, AI-powered computational methods present an efficient, cost-effective