English title:
Bayesian data analysis in the life sciences
Course ID:
790388
ECTS credits:
4,0
Title in native language:
Bayesian data analysis in the life sciences (in Eng.)
Bayesian data analysis in the life sciences (in Eng.)
Term Semester:
Autumn/Winter
Instruction language(s):
English
Course content:
Introduction to Bayesian inference with applications in the life sciences.
I Theory
1) Crash course in Machine Learning
2) Bayesian concepts - from data to models
3) Techniques for Bayesian inference with focus on variational Bayes in conditional exponential family graphical models.
4) Tools for Bayesian inference with a focus on the Python 3 library BayesPy
5) Bayesian models with applications to computational biology
II Practical applications
Based on the skills acquired in the theory part of the lecture, students will develop Bayesian graphical models to solve challenges in the life science domain. After an introduction to BayesPy, students will work on these challenges independently. A typical challenge could for example require analyzing a public data set for low level molecular biological leads and map these leads to high level representations like GO:biological process terms or implied pathways. The solutions to these mini projects will be developed in BayesPy. Students may work on their own or team up with one colleague to share work and credit. An important aspect in applied data analysis is to document all analysis steps and results. Course participants are to this end required to hand in a lab protocol which consists of BayesPy analysis scripts and results. Detailed working instructions will be handed out in the course of the lecture.
I Theory
1) Crash course in Machine Learning
2) Bayesian concepts - from data to models
3) Techniques for Bayesian inference with focus on variational Bayes in conditional exponential family graphical models.
4) Tools for Bayesian inference with a focus on the Python 3 library BayesPy
5) Bayesian models with applications to computational biology
II Practical applications
Based on the skills acquired in the theory part of the lecture, students will develop Bayesian graphical models to solve challenges in the life science domain. After an introduction to BayesPy, students will work on these challenges independently. A typical challenge could for example require analyzing a public data set for low level molecular biological leads and map these leads to high level representations like GO:biological process terms or implied pathways. The solutions to these mini projects will be developed in BayesPy. Students may work on their own or team up with one colleague to share work and credit. An important aspect in applied data analysis is to document all analysis steps and results. Course participants are to this end required to hand in a lab protocol which consists of BayesPy analysis scripts and results. Detailed working instructions will be handed out in the course of the lecture.
Expected previous knowledge:
Students are advised to take the "Machine Learning and Pattern Recognition for Bioinformatics" lecture before attending this course. Knowledge of mathematics (linear algebra), statistics and elementary programming skills are essential.
Learning outcomes:
Participants will be able to analyze simple Bayesian models analytically. Successful students will in addition know how to construct and infer a wide spectrum of Bayesian models with BayesPy and know how to apply such models for data analysis in the life sciences. Completion of the lecture will furthermore provide participants with the necessary background to understand scientific literature which enables them to tailor more complex models and Bayesian inference methods to challenges which emerge in scientific and R&D settings.
Teaching and learning methods:
The lecture consists of a theory and a practical part. The theory part consists of four half day lecturing units whereas the practical part consists of a two day introduction to BayesPy and a three day Bayesian inference boot camp.
The 4.0 ECTS of this lecture result from a workload which is structured as follows:
1) Face to face lecturing amounts to 1.5 ECTS
2) Completing the theoretical homework amounts to 0.5 ECTS
3) Completion of one comprehensive ML challenge as home work has to be handed in before the practical part starts and amounts to a work load of 0.5 ECTS. The objective of this practical exercise is providing students with a reminder of the contents of the "Machine learning and pattern recognition for bioinformatics" lecture.
4) Completion of the comprehensive BayesPy boot camp challenges amounts to 1.5 ECTS
The 4.0 ECTS of this lecture result from a workload which is structured as follows:
1) Face to face lecturing amounts to 1.5 ECTS
2) Completing the theoretical homework amounts to 0.5 ECTS
3) Completion of one comprehensive ML challenge as home work has to be handed in before the practical part starts and amounts to a work load of 0.5 ECTS. The objective of this practical exercise is providing students with a reminder of the contents of the "Machine learning and pattern recognition for bioinformatics" lecture.
4) Completion of the comprehensive BayesPy boot camp challenges amounts to 1.5 ECTS
Exam method:
This course has an innate exam character. Students are assessed on their performance in the following subtasks which contribute according to the respective scores.
1) 2 theoretical homework assignments each awarded with 5 points (teamwork with presentation)
2) One comprehensive ML task using Python scikit.learn and companion libraries is awarded with 10 points (individual assignments and group discussion of results)
3) 3 comprehensive BayesPy boot camp challenges each awarded with 10 points (task completion in teams of at most two students).
Grading scale: > 44 points: first grade (1); > 38 points: second grade (2); >31 points: grade three (3); >24 points: grade four (4);
The overall grades will be assigned in a final discussion session which takes place about a month after the BayesPy boot camp. Students may improve the proposed grade by opting for an additional verbal exam.
1) 2 theoretical homework assignments each awarded with 5 points (teamwork with presentation)
2) One comprehensive ML task using Python scikit.learn and companion libraries is awarded with 10 points (individual assignments and group discussion of results)
3) 3 comprehensive BayesPy boot camp challenges each awarded with 10 points (task completion in teams of at most two students).
Grading scale: > 44 points: first grade (1); > 38 points: second grade (2); >31 points: grade three (3); >24 points: grade four (4);
The overall grades will be assigned in a final discussion session which takes place about a month after the BayesPy boot camp. Students may improve the proposed grade by opting for an additional verbal exam.
Organisation: University of Natural Resources and Life Sciences Vienna
Country:
Austria
Acronym:
BOKU
ERASMUS+ code:
A WIEN03
Teaching period summer semester:
22. Feb 2021 - 30. Sep 2021
Teaching period winter semester:
13. Oct 2020 - 21. Feb 2021
University website: