Ateneo - UPO

The course will present the main methodologies and techniques to develop a decision support system in the bio-medical field. In particular, the first module will focus on systems based on the representation of explicit knowledge, while the second module For what regards the Bioinformatics module, the course will take care to present the main methodologies and techniques based on AI and ML for the analysis of data relating to biological sequences, gene expression, genomics and proteomics.

Z. Michalewicz et al., Adaptive Business Intelligence, Springer.
Robert Greenes. Clinical Decision Support. The Road to Broad Adoption 2nd Edition – 2014. ISBN: 9780128100240

Hiroshi Mamitsuka: Textbook of Machine Learning and Data Mining: with Bioinformatics Applications. ISBN-13: 978-4991044502
Pierre Baldi & Soren Brunak: Bioinformatics: The Machine Learning Approach, Second Edition.: Mit Press, 2001
Philip Compeau & Pavel Pevzner: Bioinformatics Algorithms. An active Learning Approach. Volumi I e II. Active Learning Publishers. 2018

Students must acquire the following knowledge, competences, and abilities: principal types of knowledge-based systems for supporting clinical decisions, with particular attention to the diagnosis and patient treatment; principal methodologies and techniques adopted in this field; analysis of enabling factors and of critical issues for the adoption of these systems in medical practice; methodologies and techniques for: knowledge representation and management, prediction, optimization, adaptation; ability to identify, interpret, encode and process various types of biological data for machine learning applications.
Students must also be able to apply such notions in answering theoretic questions as well as exercises, or in illustrating (orally) an in-depth study carried out independently

Database fundamentals

Direct lessons in classroom or in lab. Classroom lessons will also include example questions or exercises useful for the final test. Example tests can also be provided to students. Slides, textbooks indications and additional material will be provided also through the DIR platform. In this way, the students who do not attend will be allowed to easily follow the course progression.

The examination about the first part of the course (knowledge-based approaches) will be written, and will contain theoretical questions and, possibly, small exercises. Possibly (e.g., on student's request), the same exam type will be executed orally.

The test about the operative knowledge module will be written and composed of three questions at least, focused on different course topics chosen among operative knowledge management, prediction, optimization, adaptation. The final score will take into account the partial scores of the different questions (e.g., calculating the average). Possibly (e.g., on student's request), the same exam type will be executed orally.
For the bioinformatics part, the student will have to present an in-depth study in the form of a seminar on a topic related to the topics covered.

The module dedicated to knowlege-based CDSS will face the following aspects: CDSS architectures and generalities, knowledge sources and knowledge acquisition, decision rules, diagnostic systems, computer-interpretable guideline systems, ontologies and vocabularies, systems for the management of clinical trials, analysis of critical and enabling factors for the practical applicability of CDSS.

The operative knowledge module will be organised in four parts: 1. operative knowledge management: Case-based Reasoning; advanced Case-based Reasoning solutions (time series data, fuzzy CBR, process-oriented CBR); 2. prediction: classification, regression, time series; mathematical models, distance-based models, logic models, heuristic models; hybrid methods; 3. optimisation: mathematical methods, evolutionary algorithms, ant systems, hybrid methods; 4. adaptation: techniques to improve the prediction module performances.
For the bioinformatics module the following topics will be covered: Introduction to cellular and computational biology. Essential bioinformatics skills (databases, APIs, frameworks). Feature Engineering. Data imputation, Dimensionality Reduction, Applications to linear and logistic regression methodologies to biological data. Decision Trees, Random Forest, and eXtreme Gradient Boosting for biological data. Use of Hidden Markov Models for the identification of protein coding genes. Examples of Deep Learning-based approaches to analyzing biological data.

Knowledge: The student must have acquired: the notion relating to the main types of knowledge-based systems for supporting clinical decisions, for the diagnosis and/or patients treatment, to the methodologies and techniques adopted in these fields. Analysis of enabling factors and critical issues for the adoption of these systems in medical practice; the notions of: knowledge management, prediction, optimisation, adaptation; the knowledge to understand and interpret the various problems and the various types of biological data presented during the course. They must also be able to demonstrate the understanding and the acquisition of the operating principles of the various methodologies and techniques presented.
Competence and ability: The student must have acquired the ability to evaluate the suitability of the techniques and methodologies developed in the course for the management of real-world case studies. Students must demonstrate the ability to critically identify the best analysis approach, based on the relative strengths and weaknesses, to be applied to different types of biological data.
Judgement autonomy: ability to work in an autonomous way, and deal with complex situations or with incomplete information
Communication abilities: ability to clearly describe the methodologies and techniques studied, and their applications, to both expert and non-technical users; ability to justify design or implementation choices and clearly communicate them also to a non-expert audience; ability to clearly describe the operating principles of the studied methodologies, the possible advantages and disadvantages and their applicability to real problems to different types of audiences with different levels of experience.

Learning capability: Students must have developed adequate knowledge of basic methodologies and the ability to analyze the main characteristics of clinical decision support systems and of the proposed methodologies for the analysis of biological data, to be able to learn, evaluate and/or develop new systems. Students must be able to describe and properly choose the correct techniques to realize an healthcare decision support system. Students must have acquired the operating principles of the methodologies analyzed to propose changes and adaptations for improving their applicability to the specific case. Identify and critically evaluate scientific information from publications in this field