CoChat

Machine Learning Methods in Drug Discovery

Open access

The DOI matches its Crossref record; 103 of the 103 references it cites were found in a registry or an index.

Abstract

The advancements of information technology and related processing techniques have created a fertile base for progress in many scientific fields and industries. In the fields of drug discovery and development, machine learning techniques have been used for the development of novel drug candidates. The methods for designing drug targets and novel drug discovery now routinely combine machine learning and deep learning algorithms to enhance the efficiency, efficacy, and quality of developed outputs. The generation and incorporation of big data, through technologies such as high-throughput screening and high through-put computational analysis of databases used for both lead and target discovery, has increased the reliability of the machine learning and deep learning incorporated techniques. The use of these virtual screening and encompassing online information has also been highlighted in developing lead synthesis pathways. In this review, machine learning and deep learning algorithms utilized in drug discovery and associated techniques will be discussed. The applications that produce promising results and methods will be reviewed.

Overview

The claims and statistics of this paper haven’t been checked yet.

Free: it reads the open-access text and checks the first 5 claims that cite a source we can check. Create a free account to check every claim.

A few free checks a day · papers already verified open instantly · results are public

Integrity signals

Registry

Crossref · 10.3390/molecules25225277

Cites

103 of 103 references verified0 citation errors · 0 not checkable

Statistics

Figures

Figure 1 The general steps in drug discovery. Machine learning (ML) and deep learning (DL) algorithms may participate in each of the four steps listed, e.g., by mining proteomic in target discovery, discovering small molecules as candidates in lead discovery, developing quantitative structure-activity relationship models to optimize lead structures for improved bioactivity, and analyzing massive assay results.
Figure 1 The general steps in drug discovery. Machine learning (ML) and deep learning (DL) algorithms may participate in each of the four steps listed, e.g., by mining proteomic in target discovery, discovering small molecules as candidates in lead discovery, developing quantitative structure-activity relationship models to optimize lead structures for improved bioactivity, and analyzing massive assay results.CC BY · PubMed Central
Figure 2 Schematic view of drug development using random forest (RF) (a) and support vector machine (SVM) (b). (a) RF reaches the final decision of drugs by combining the results of randomly-created decision trees (three trees are shown for simplicity). There are multiple features that the computational queries look for in both target and drug. When there is a compatibility match, it proceeds to the next step to match additional features. A series of datasets is inputted into the query, and each tree is responsible for computing a prediction. The prediction picked by most trees is used for the next step. The system of using many decision trees is intended to minimize errors mathematically. (b) SVM utilizes similarities between the classes, called support vectors, to distinguish between the classes based on the trained features. It formulates hyperplanes that separate two classes (can be multiclass, if needed). SVM incorporates multiple training sets depending on the classifiers and formulates compounds’ status (active or inactive). During the process, compounds are separated into three sections: Non-selective compounds (active), selective compounds (active), and in the margin are inactive compounds. Although non-selective compounds are active, they are not selective towards the protein of interest. In contrast, selective compounds are active and selective towards the protein of interest.
Figure 2 Schematic view of drug development using random forest (RF) (a) and support vector machine (SVM) (b). (a) RF reaches the final decision of drugs by combining the results of randomly-created decision trees (three trees are shown for simplicity). There are multiple features that the computational queries look for in both target and drug. When there is a compatibility match, it proceeds to the next step to match additional features. A series of datasets is inputted into the query, and each tree is responsible for computing a prediction. The prediction picked by most trees is used for the next step. The system of using many decision trees is intended to minimize errors mathematically. (b) SVM utilizes similarities between the classes, called support vectors, to distinguish between the classes based on the trained features. It formulates hyperplanes that separate two classes (can be multiclass, if needed). SVM incorporates multiple training sets depending on the classifiers and formulates compounds’ status (active or inactive). During the process, compounds are separated into three sections: Non-selective compounds (active), selective compounds (active), and in the margin are inactive compounds. Although non-selective compounds are active, they are not selective towards the protein of interest. In contrast, selective compounds are active and selective towards the protein of interest.CC BY · PubMed Central
Figure 3 The general scheme of deep neural network (DNN) (a) and recurrent neural network (RNN) (b). (a) DNN consists of an input layer followed by several hidden layers and an output layer. In this case, the input layer utilizes feature vectors generated by a convolutional network. The progression of the NN follows a single path through hidden layer 1 (HL1) to HLn, indicating the feedforward nature of the NN. The generated outputs are often processed using supervised learning techniques for the identification and collection of sensible interactions. (b) RNN begins with a seed, S, which is inputted into the system. Through the use of algorithmic processing, the seed is turned into a reference vector, V1, which is used by the HL to generate a vector output, V2. V2 is subsequently optimized through input training sets and creates the output, O. The generation of these outputs eventually leads to the creation of a gatherable data set. In the meantime, the HLs feed forward to provide information from previous steps. One example is chemical structure generation using SMILE string characters as seeds; hence the desired gathered outputs would be a string of SMILE characters that would be the desired structure. The dataset created in the figure is gathered and analyzed into the resultant molecules.
Figure 3 The general scheme of deep neural network (DNN) (a) and recurrent neural network (RNN) (b). (a) DNN consists of an input layer followed by several hidden layers and an output layer. In this case, the input layer utilizes feature vectors generated by a convolutional network. The progression of the NN follows a single path through hidden layer 1 (HL1) to HLn, indicating the feedforward nature of the NN. The generated outputs are often processed using supervised learning techniques for the identification and collection of sensible interactions. (b) RNN begins with a seed, S, which is inputted into the system. Through the use of algorithmic processing, the seed is turned into a reference vector, V1, which is used by the HL to generate a vector output, V2. V2 is subsequently optimized through input training sets and creates the output, O. The generation of these outputs eventually leads to the creation of a gatherable data set. In the meantime, the HLs feed forward to provide information from previous steps. One example is chemical structure generation using SMILE string characters as seeds; hence the desired gathered outputs would be a string of SMILE characters that would be the desired structure. The dataset created in the figure is gathered and analyzed into the resultant molecules.CC BY · PubMed Central

Verification

Claims the paper makes

The claims and statistics of this paper haven’t been checked yet.

Free: it reads the open-access text and checks the first 5 claims that cite a source we can check. Create a free account to check every claim.

A few free checks a day · papers already verified open instantly · results are public

Claims other papers make about this one

Each is checked against this paper’s own text: does it say what the citing paper says it does?

Self-check

Percentages, sums and reported differences re-computed from the paper’s own numbers.

The self-check needs the paper’s full text.

Citations

Cites

Cited by

See all 460

Metadata

Authors
Patel, L., Shukla, T., Huang, X., Ussery, D. W., & Wang, S.
Affiliations
  • University of Arkansas at Little Rock (US)
  • Arkansas State University (US)
  • University of Arkansas for Medical Sciences (US)
Corresponding author
Wang, Shanzhi
Venue
Molecules
Publisher
MDPI AG
Published
2020-11-12
Volume
25
Issue
22
Pages
5277
ISSN
1420-3049 (ISSN-L)
OpenAlex
W3099252273
Cited by
460
References
103
Citations by year
2026: 58 · 2025: 93 · 2024: 104 · 2023: 92 · 2022: 84 · 2021: 27
FWCI
4.48
Type
journal article
Language
en
Open access
Locations
  • Molecules (publishedVersion, cc-by)
  • PubMed Central (publishedVersion, cc-by)
  • PubMed (publishedVersion)
  • DOAJ (DOAJ: Directory of Open Access Journals) (submittedVersion, cc-by-sa)
Funders
  • The Arkansas Department of Education (ADE) (2019SURF)
  • Federal TRIO Programs, United States Department of Education (McNair (LP))
  • NSF (1946391)
Topics
  • Life Sciences › Pharmacology, Toxicology and Pharmaceutics › Drug Discovery › Computational Drug Discovery Methods
  • Physical Sciences › Materials Science › Materials Chemistry › Machine Learning in Materials Science
  • Life Sciences › Biochemistry, Genetics and Molecular Biology › Molecular Biology › Protein Degradation and Inhibitors
Keywords
machine learning in drug discovery, deep learning, drug target identification, virtual screening, lead compound identification, high-throughput screening, big data, drug candidate design
MeSH terms
Machine Learning, Deep Learning, Algorithms, Bayes Theorem, Humans, Monte Carlo Method, Software, Reproducibility of Results, Databases, Factual, Computational Biology, Internet, Drug Discovery, Support Vector Machine

About these checks

Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.

Statistics and arithmetic are recomputed, not judged.

Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.

What counts as a claim ›How we measure confidence ›How we check this ›

Are you an author, or do you see a problem with this page? Request a review

Data sources

Bibliographic data: Semantic Scholar, OpenAlex, Crossref.

Contains information from the Semantic Scholar Academic Graph Datasets, made available under the ODC Attribution License (ODC-BY 1.0).

Semantic Scholar’s data is described in: Kinney, R., et al. (2023). The Semantic Scholar Open Data Platform. arXiv:2301.10140.