International Research Journal of Engineering and Technology (IRJET)
e-ISSN: 2395-0056
Volume: 11 Issue: 12 | Dec 2024
p-ISSN: 2395-0072
www.irjet.net
Deep Learning in Drug Discovery and Pharmaceutical Research Kallesh S C1, Siddesh D S2, Thejaswi M R3, Kushi D N4 1Bachelor of Engineering, Information Science and Engineering, Bapuji Institute of Engineering and technology,
Karnataka, India[1,2]
2Bachelor of Engineering, Information Science and Engineering, Bapuji Institute of Engineering and technology,
Karnataka, India [3,4] ---------------------------------------------------------------------***--------------------------------------------------------------------advanced chemocentric machine-learning methods with a Abstract - Deep learning has revolutionized the field of focus on emerging “deep learning” concepts. We highlight recent advances in the field and point to prospective applications and developments of this potentially game changing technology for drug discovery. Ageneral task for machine-learning is to uncover the relationship between the molecular descriptors used and the measured activity of the compounds to obtain qualitative classifiers or quantitative structure-activity relationship (QSAR) models. Feature extraction from the descriptor patterns is the decisive step in the model development process.[6,7] In current cheminformatics applications, the prevalent machinelearning architectures are “shallow” and contain a single layer of feature transformation. These architectures include linear and nonlinear principle component analysis, k-means clustering methods, partial least square projection to latent structures, decision trees, multivariate linear regression, linear discriminant analysis, support vector machines (SVMs), logistic and kernel regression, multi-layer Perceptrons and related neural network approaches.[8] Although all of these methods have proven to be useful for (Q)SAR modeling and molecular design,[9] the single feature transformation step into a suitable space for the subsequent application of a linear pattern separation model might limit their modeling and representational power when applied to more complex data and setups. A reason for their success in pharmacological applications may stem from the fact that a major part of the complexity inherent to molecular interactions has been engineered into the descriptors employed as patterns for model training, thereby allowing single layer machinelearning architectures to tackle the problem.[10] One challenging question is whether the underlying data complexity and hidden features can be more efficiently dealt with by shifting attention from descriptor engineering to the architecture of the machine-learning system and the training of the algorithms involved. This is the domain of “deep learning”.
drug discovery and pharmaceutical research. This paper provides a comprehensive review of the applications of deep learning in drug discovery and pharmaceutical research. We discuss the various deep learning architectures and techniques used in drug discovery, including convolutional neural networks (CNNs), recurrent neural networks (RNNs), and generative adversarial networks (GANs). We also highlight the challenges and limitations of using deep learning in drug discovery and pharmaceutical research. Key Words: bioinformatics · cheminformatics · drug design · machine-learning · neural network · virtual screening
1.INTRODUCTION Machine-learning provides a theoretical framework for the discovery and prioritization of bioactive compounds with desired pharmacological effects and their optimization as drug-like leads. Biological target identification and protein design are emerging areas of application. Among the many machine-learning approaches in molecular informatics, chemocentric methods have found widespread application. Their underlying logic typically follows three steps. First, there is the selection of a problem-specific set of descriptors that are believed to capture the essential properties of the molecules involved. At present, there are over 5000 diverse molecular representations (“descriptors”) that address the various properties of chemical entities.[1] Second, a metric or scoring scheme is used to compare the encoded molecules to one another.[2] Finally, a machine-learning algorithm is employed to identify the features that may serve to qualitatively or quantitatively distinguish the active from the inactive compounds.[3] Artificial neural networks (ANNs) were among the first methods borrowed from the computer sciences for this purpose. In 2013, public attention was drawn to a multiproblem QSAR machine-learningchallengein drug discovery posted by Merck. This competition on drug property and activity prediction was won by a deep learning network with a relative accuracy improvement of approximately 14% over Merck’s in-house systems and resulted in an article in The New York Times. [5] Here, we present state-of-the-art of
© 2024, IRJET
|
Impact Factor value: 8.315
|
ISO 9001:2008 Certified Journal
|
Page 585