99486 - Simulation Methods and Machine Learning in Medicinal Chemistry

Learning outcomes

At the end of the course, the student knows the basic aspects of consolidated and emerging computational approaches and methodologies in the pharmaceutical chemistry field. The student is able to address the many problems that characterize the phases of drug discovery and development through the choice of the most appropriate computational tool.

Course contents

• Introduction to the course.
  Distinction between simulation methods and Machine Learning in the field of medicinal chemistry. Distinction between supervised and unsupervised Machine Learning.

• Review of concepts learned in previous courses.
  Discrete and continuous variables. Empirical distributions. Univariate and multivariate statistical analysis. Operations on vectors and matrices. Eigenvectors and eigenvalues of a matrix. Python exercises.
  

Machine Learning methods:

• The concept of "molecular featurization".
  Molecular fingerprints. Topological descriptors. Linear notations. Python exercises.

• Simple and multiple linear regression.
  Loss functions. The training process. The method of least squares. Numerical methods. Evaluation of the performance of a regression model. Python exercises.

• Feature selection and feature extraction.
  Univariate linear filters. Wrapper methods. Regularization methods. Principal Component Analysis. Python exercises.
  

• Neural networks.
  The Multi-layer Perceptron. Training and optimization. Regularization and hyperparameter tuning. Introduction to generative methods. Autoencoders and variations. Prediction of pharmaceutically relevant molecular properties using neural networks. Python exercises.

Methods based on molecular simulations:

• Introduction to simulation methods.
  The drug action process. Drug-target interaction: thermodynamics and kinetics aspects. Review of intermolecular interactions, enthalpic and entropic contributions.

• Introduction to statistical mechanics.
  Microscopic definition of entropy. Microstates and macrostates. Boltzmann probability distribution. Helmholtz free energy.

• Molecular mechanics.
  Force fields. Minimization algorithms.

• Sampling methods.
  Introduction to the Metropolis Monte Carlo method. Introduction to the Molecular Dynamics method. Integration of the equations of motion. Brief overview of temperature control algorithms (thermostats). Boundary conditions. Trajectory analysis. Prediction of experimental observables of pharmaceutical interest using Molecular Dynamics simulations: binding free energy and association/dissociation rate constants.

Readings/Bibliography

Scientific articles and reviews suggested by the teacher.

Teaching methods

Frontal lectures and Pyhton exercises.

Assessment methods

• Students may choose between the following examination modalities:

1. Oral presentation and discussion of a scientific article selected from the international literature and related to one of the topics covered in the course. The presentation will be conducted in a journal club format, with all students opting for this examination modality in attendance. This format is intended to encourage discussion and stimulate questions from participants, thereby fostering critical thinking and the ability to critically analyze the scientific literature.
2. Presentation of an original research project related to one of the topics covered in the course. For projects focused on machine learning methodologies, the work may be presented in the form of a Jupyter Notebook and should include the implementation and application of the methods studied to a set of model molecules selected by the student.

• Regarding the assessment of learning outcomes, limited, declared, and non-substantial use of AI is permitted for support activities (e.g., summarization and rewording). Substantial use of AI to complete any part of the assessment is not permitted.

• Students with learning disorders and\or temporary or permanent disabilities: please, contact the office responsible as soon as possible so that they can propose acceptable adjustments. The request for adaptation must be submitted in advance (15 days before the exam date) to the lecturer, who will assess the appropriateness of the adjustments, taking into account the teaching objectives.

Teaching tools

Slides, scientific publications and other teaching material made available through the Virtual Learning Environment platform.

Office hours

See the website of Matteo Masetti