Natália Cordeiro

[introductory/intermediate] Multi-Tasking Machine Learning in Drug and Materials Design

Summary

Quantitative Structure-Activity Relationships (QSAR) modelling is a well-known computational technique with wide applications in several fields such as drug or materials design, toxicity predictions, etc. This series of lectures will start with an introduction to the basics of QSAR modelling, and then the focus will be on a more advanced methodology that allows mining the response data coming from multiple conditions, the so-called multitasking QSAR (mt-QSAR), showing its relevance. A number of different machine learning techniques can be used for setting up mt-QSAR models, in which artificial neural network (ANN) techniques, especially multilayer perceptron (MLP) or deep neural networks (DNN), play a significant role. Throughout the lectures, case studies on how ANN methodologies can be useful in developing validated and predictive mt-QSAR models will be given, plus comparisons to other machine learning techniques. The lectures will also include hand-on applications of QSAR-Co-X, an open access Python based tool that has been developed for supporting ANN (and other machine learning techniques) based on mt-QSAR modelling.

Syllabus

• Introduction to Quantitative Structure-Activity Relationships (QSAR).
• Including multiple conditions: Multitasking QSAR (mt-QSAR) modelling.
• Basics on supervised learning approaches applied to mt-QSAR.
• Hands-on tutorial: Implementation of QSAR-Co-X (an open access Python based tool) for developing mt-QSAR models with ANN and other machine learning tools.

References

Halder AK, Moura AS, Cordeiro MNDS (2022). Moving Average-Based Multitasking In Silico Classification Modeling: Where Do We Stand and What Is Next? Int J Mol Sci. 23, 4937. https://doi.org/10.3390/ijms23094937

Ambure P, Halder AK, Díaz HG, Cordeiro MNDS (2019). QSAR-Co: An Open Source Software for Developing Robust Multitasking or Multitarget Classification-Based QSAR Models. J Chem Inform Model 59, 2538−2544. https://doi.org/10.1021/acs.jcim.9b00295

Halder AK, Cordeiro MNDS (2021). QSAR-Co-X: an open source toolkit for multitarget QSAR modelling. J Cheminform 13, 29. https://doi.org/10.1186/s13321-021-00508-0

Halder AK, Cordeiro MNDS (2021). Multi-Target In Silico Prediction of Inhibitors for Mitogen-Activated Protein Kinase-Interacting Kinases. Biomolecules, 11, 1670. https://doi.org/10.3390/biom11111670

More references will be provided directly during the tutorial and in the slides.

Pre-requisites

Basic knowledge in mathematics and machine learning. Basic programming abilities in Python. (Anaconda must be installed in the computers.)

Short bio

M. Natália D.S. Cordeiro is an Associate Professor of Theoretical Chemistry at the University of Porto (Portugal). Since 2003, she heads the Cheminformatics and Materials research group (https://www.fc.up.pt/quimat/) at the Associated Laboratory for Green Chemistry (https://laqv.requimte.pt/). The research of her group focuses on the computational modelling of novel functional materials, their physical and chemical properties, targeting their application in catalysis, electrochemistry, water purification systems, sensors, drugs, and so forth, as well as probing their potential safety. Equally varied as the research topics are the methods employed to study them, which involve molecular simulations as quantum-mechanical calculations and machine learning tools.