**3. Artificial intelligence (AI) and machine learning (ML) technologies in drug discovery, diagnosis, and health care of COVID-19**

#### **3.1 Drug discovery**

Therapeutics: AI and ML in treatment discovery development and/or drug repurposing for COVID-19 based on:


AI may contribute to the advancement of resources to support doctors and ultimately enhance medical outcomes. Fuzzy logic can be used in decision support systems to replicate patient decision-making processes [80–82]. Admittedly, machine learning applied is to clinical data that are regularly collected will produce new knowledge and potentially new perspectives that clinicians lack.

Drug repurposing is hoped to offer a way to establish COVID-19 avoidance and cure policies. For instance, the researchers built a DL approach to classifying current *Perspective Chapter: Repurposing Natural Products to Target COVID-19 – Molecular Targets... DOI: http://dx.doi.org/10.5772/intechopen.103153*

and mercantile medicines for "drug-repurposing," i.e., identifying a quick treatment using existing medicines that can be introduced to patients immediately. The idea that recently created treatment typically needs years to succeed is reviewed before getting to the public motivates research. Although the results are not accepted clinically, new approaches to combat COVID-19 disease are already opening up [83]. In silico medicine is suggested in [84] using the deep generative model to explore drugs (identifying new medicines). This analysis may be used for simulations and computer modeling to obtain compounds for COVID-19 coronavirus by new molecular entities.

IBM reported that it is now offering an analysis service based on the cloud using the COVID-19 dataset that has been educated [85]. Besides, IBM has implemented its proposed drug discovery AI technology, in which 3000 novel COVID-19 molecules have been produced [86]. In the year 2020, a systematic analysis was developed by Zeng et al. [87] to find drugs for COVID-19. With the support of active Amazon Web Services (AWS), a DL-based model was developed, and 41 data on drug types were validated. As for performance metrics, true-positive rate (TPR), false-positive rate (FPR), etc., have been presented, and the approach suggested by the author is explicit that DL serves as an important instrument for exploring therapeutics.

#### **3.2 Diagnostics**

Earlier, our research team had presented the usefulness of AI and ML in diagnosis of several diseases [88–90]. However, COVID-19 diagnosis was based on AI.


#### *3.2.1 Image data and deep learning*

Nour et al. [91] have developed a DL model for COVID-19 detection, as CNN is applied as a feature extractor. For performance assessment, chest X-ray images dataset is taken into account. For feeding ML methods such as K-nearest neighbour (KNN), Decision Tree (DT), and *support vector machines* (SVM), the deep feature that has been extracted with the aid of CNN is utilized. Precision, F-score, etc., are used as output variables. Among other suggested approaches, SVM yields greater precision.

Pereira et al. [92] proposed a new model for forecasting the dynamics of COVID-19 that have cases that have happened in other countries or places with similar emission patterns. For all subregions and accessible countries, they implemented a grouping algorithm.

#### **3.3 Health care**

Big data in the administration of hospitals, epidemiology, insurance, medication interactions and complications, outcomes reviews based on quality, epidemic tracking.

Speech datasets include breath sounds and cough, which can be utilized for COVID-19 diagnoses and its prediction for illness seriousness. Machine learning, statistical techniques, and big data may be used to the datasets for prediction functions about the disease. Various open-source datasets for COVID-19 included mobility, diagnosis, contagion assessment, NPI analysis, statistic relationships, and sentiment analysis.
