used in combination with computational de novo technology, it assists in extracting

data from previously defined compounds, allowing for the development of chemi-

cally accurate and biologically active compound structures (Oskooei et al. 2018)

(Fig. 1.2).

•

One of the advanced technologies is the recurrent neural networks (RNNs). RNN

produces the new character strings corresponding to molecules within the chemi-

cal space. In this way, many more drug-like molecules can be synthesized,

matched with the drug target information to be placed in a particular region of

chemical space of drugs (Eureka 2019a).

•

An open-source software, the “CellProfiler” can quantitatively measure the

phenotypes from thousands of images by automatically recognizing the cells

and measuring their properties in the image (Steensberg and Simons 2015). The

phenotype of cells is then recorded using “CellProfiler Analyst.” For example, a

healthy or a diseased cell can be taken to compare their morphology from a

patient. Their profile difference can be used as a diagnostic tool (Eureka 2019b).

Other open-source applications, such as PaccMann, INtERAcT, and PIMKL from

IBM Research, Zurich (Switzerland), are also available these days (Manica and

Cadow

2019). PaccMann’s sensitivity

of cancer

cells

is

predicted

by

incorporating transcriptomics, cellular protein interactions, and compound molec-

ular structure (Oskooei et al. 2018). Likewise, INtERAcT uses unsupervised

machine learning to scrutinize cancer research publications and draw interactions,

such as protein-protein interactions. Similarly, a machine-learning algorithm

PIMKL is used to infer phenotype from multi-omics data.

•

Pharmacovigilance (drug safety science) is the science of collecting, detecting,

assessing, monitoring, and preventing adverse drug reactions (ADRs). Since there

is now such a massive amount of data available, AI and machine learning will

enhance the above processes. Due to the expanded compilation of electronic

health records (EHRs) and access to freely accessible resources, the use of AI

approaches for pharmacovigilance is growing day by day. Machine learning

(ML) and deep learning (DL) techniques are now being used to replace conven-

tional strategies, such as quantitative structure-activity relationships (QSAR) for

determining preclinical safety (Kantarjian et al. 2012).

Fig. 1.2 Artificial intelligence in drug discovery

1

Emerging Technologies: Gateway to Understand Molecular Insight of. . .

9