Teaching Responsibility

LJMU Schools involved in Delivery:

Pharmacy & Biomolecular Sciences

Learning Methods

Lecture

Practical

Tutorial

Workshop

Module Offerings

7118PHASCI-JAN-CTY

Aims

To enable students to compare the advantages, disadvantages and applications of advanced computational modelling approaches, considering metrics such as adherence to OECD principles, applicability domain, reproducibility, transparency and statistical performance. To equip students with the skills necessary to build, optimise, interpret and report quantitative structure-activity relationship models.

Learning Outcomes

1.
Explain the principles of good modelling practice, particularly in relation to OECD principles and published in silico toxicology protocols.
2.
Critically compare a range of advanced computational methods used in predictive toxicology.
3.
Interpret existing models and explain the key (statistical) information associated with different types of models
4.
Build, optimise, explain and report quantitative structure-activity relationship (QSAR) models
5.
Demonstrate proficiency in the use of a range of computational tools for QSAR model building.

Module Content

Outline Syllabus:Building quantitative structure-activity models, using appropriate software (such as Minitab) with reference to good modelling practice - OECD Principles/ published in silico toxicology protocols. Identifying appropriate endpoint data, descriptors and statistical approaches (e.g. use/interpretation of multilinear regression, r2, rCV2, Q2, outliers, confusion matrices, false positives and negatives, Matthews correlation coefficient, discriminant functions etc.) Pipeline environments for model building; exemplar models (e.g. VEGA; qsardb.org) Interpreting and assessing the usability (for a given purpose) and repeatability of existing QSARs. Model reporting, evaluation and validation: documentation, QSAR model reporting format, Additionally, examples will be provided of more complex variable selection and model building methods – these will be updated according to developments in the area but may include: Genetic algorithms, artificial neural networks, support vector machines, deep learning methodology, random forests etc. Examples of where the methods have been applied, pros and cons of the approaches and comparison of “black box” versus transparent methods in predictive toxicology.
Module Overview:
Offers a practical introduction to more advanced, state-of-the-art methods to predict chemical toxicity, including how the methods are applied and assessed.
Additional Information:The contents of this module link directly with module 7117PHASCI (Computational Methods I: Data and modelling). This module will follow logically from the preceding module enabling greater depth of exploration of more complex modelling techniques and state-of-the-art methods. The content will evolve with the latest developments in the area.

Assessments

Portfolio

Centralised Exam