Artificial learning and machine learning are playing pivotal roles in the society, especially in the field of medicinal chemistry and drug discovery. Particularly its algorithms, neural networks, or other recurrent networks drive this area. In this review, the authors have taken into account the diverse use of AI in a number of pharmaceutical industries including discovery of drugs, repurposing, development of pharmaceutical drugs, and clinical trials. In addition, the efficiency of these artificial or machine learning programs in achieving the target drugs in short time period along with accurate dosage and cost of the drug have also been discussed. Numerous applications of AI in property prediction such as ADMET have been used for prediction of strength of this technology in QSAR. In case of de-novo synthesis, it results in generation of novel drug molecules with unique design making this a promising field for drug design. Moreover, its involvement in synthetic planning, ease of synthesis, and much more contribute to automated drug discovery in the near future.
TopIntroduction
It has been estimated that an average cost of bringing drug to the marketplace is found to exceed about 3 billion dollars. This increment in cost is attributed to the two factors which include, the number of clinical trials resulting in attraction to about 80-85% as required for approving drugs for humans. Secondly, the complexity of the discovery phase of drugs demanding a considerable amount of investment for both time and resources (Díaz, Dalton, & Giraldo, 2019). A strong pipeline of candidate drug in preclinical trials will in turn have significant downstream effects in consideration of total approval. The advancement in both the computer software’s and in-vitro approaches aim to employ as well as improve the various other aspects of drug discovery cycle and the test refered as quitensial design make test analyse DMTA. The area of increasing interest is utilization of data driven synthetic tools with the objective to reduce the number of failures with the subsequent increase in output of drugs during synthesis of novel molecular drug subunits. History from 1960 shows that computer aided synthesis planning CASP when the Corey group first disclosed LASHA which presents a rule based approach for retrosynthetic planning. This publication was the prime key for providing definition to heuristics involving chemical synthesis which could be a valuable tool for the software involving synthesis planning of drugs. From 1960s to 1990, many groups disclosed that the advancement in computer based planning of synthesis were limited by computation resources and rely wholly on human based rules for reaction (Yang, Wang, Byrne, Schneider, & Yang, 2019).
The early progenitors provide the basis for many of the commercial software as for example Synthia formally termed as chematica. In addition to ICSynth where the rules coded by hand are utilized in addition to following guidelines for heuristics in relation to negative synthetic pathways.
From the past few decades, more automated methods have been found regarding synthesis process which use the subset of an artificial learning method being referred as machine learning for inferring the reactivity from previously available data which has provided a visible alternative to expert rule-based algorithm. Hence both the expert based and ML can come under the umbrella niche of AI approach. The first one using the information from crafted knowledge and second presents the example of using statistical learning method. Each of them has its own specific advantage for drug synthesis planning process. But the machine learning has extended to incorporate further new reactions because they are published based on extraction or training pipelines which in turn reduces burden on experts belonging to this category i.e. researchers etc. As much as reactions operate in the industry or company, the automated method provide prediction of the candidate molecule and results more robustly(Chan, Shan, Dahoun, Vogel, & Yuan, 2019).
Both the rule based and machine-based learning has provided valuable tool for planning of synthetic route being executed in laboratory and evaluated by chemist during research. As for example Synthia has developed route for the medically related compounds which is far better than the routes developed by experts. As Seiger et al. expressed that their researchers have not preferred the last or previous route developed by literature but has taken notes from their novel algorithm based route in double blind evolution process in double blind evolution process. Automated platforms are coupled to synthesis planning tools for a varying level of human intervention. However this field is in its early stages to use CASP or fully automated planning, this resulting in initial success which provides the better tool for the drug development process following DMTA cycle (Jing, Bian, Hu, Wang, & Xie, 2018).