Algorithms in Bioinformatics - #22125

Information for participants

GENERAL SCHEDULE
Lectures will be in the morning from 9.00 - 12.00, and exercises in the afternoon from 13.00 - 17.00.

The morning sessions will consist of lectures and small practical exercises introducing the different algorithms, and the afternoon sessions will consist of programming exercises where the algorithms will be implemented.

The main programming language will be Python and all program templates provided in the course will be written in Python. Prior detailed knowlegde of Python programming is NOT required but will make following the course very much easier. However, basic programming skills are required to follow the course.

PROGRAMS AND TOOLS

To do the exercises you must have a Linux/Darwin envirronment on your laptop.
You must have a Python installation and be able to run "jupyter-notebook" in this Linux/Darwin envirronment. It is hence not an option to only run python from your Windows installation.
If you work on a Windows machine, one solution to this is described here Installing_Ubuntu_WSL_windows.
For other solutions and more help refer to Peter Wads notes on Course preparation from 22113/22163 - Unix & Python Programming for Bioinformaticians and Install Jupyter Notebook. Here, you can also find information on how to install Anaconda Python in your WSL2. Joakim (TA on our course) has added a few slides on how to start Jupyter Notebook after having installed Anaconda in WSL in the PowerPoint file linked here 22113_01-AnacondaInstall.ppt
Python for Non-Programmers
Python Numpy Tutorial

Course Programme
Please note that the programme is updated on a regular basis - click the 'refresh' button once in a while to make sure that you have the most updated information

LITERATURE:

The course curriculum consists of review paper and selected chapters from Immunological Bioinformatics, Lund et al., MIT Press, 2005. All course material will be made available online during the course. All course material is available either as open access or here Course material

Wednesday, 4. June

Introduction to course. PSSM construction from pre-aligned sequences including pseudo count. Python recap
Morten Nielsen

BACKGROUND TEXTS Essentials:

Additionals:

9.00 - 9.15: Introduction to course; Introduction to course [PDF] (MP4).
9.15 - 9.30: Introduction to the immune system [PDF] (MP4).
9.30 - 11.20 (coffee break included): Weight matrix construction[PDF]. [PPTX] Weight matrix construction (MP4).; Logo Handout; Answers; Handout. Estimation of pseudo counts; Answer
11.20 - 11.40: Some notes on sequence alignment [PDF] (MP4)
11.40 - 12.00: Questions to the mornings lectures and other general issues; Checking that we all have python and jupyter-notebook installed and running
12.00 - 13.00: Lunch
13.00 - 17.00: A brief introduction to Python programming and Jupyter-notebooks; Python intro; Python Answers; Implementation of PSSM construction from pre-aligned sequences including pseudo count correction for low counts and sequence clustering; PSSM construction and evaluation; PSSM answers

Thursday, 5. June. Holiday

Friday, 6. June

Sequence alignment, Dynamic programming, and Psi-Blast
Morten Nielsen

BACKGROUND TEXTS
Essentials:

Additionals:

9.00 - 9.30: Questions to yesterdays lectures and exercises
9.30 - 11.00: Blosum matrices [PDF] (MP4); Sequence alignment [PDF] . [PPTX] . (MP4); Handout (O3); Handout (O2); Handout answers
11.00 - 12.00: Blast alignment heuristics, Psi-Blast, and sequence profiles [PDF] . [PPTX] .; Psi-Blast handout.
12.00 - 13.00: Lunch
13.00 - 17.00: Implementation of the Smith-Waterman Dynamic programming algorithm; Matrix dumps from alignment programs (to be used for debugging); Answers to sequence alignment exercise

Monday, 9. June. Holiday

Tuesday, 10. June

Data redundancy reduction algortihms
Optimizations methods
Gibbs sampling
Morten Nielsen

BACKGROUND TEXTS - Data redundancy reduction and Gibbs sampling
Essentials:

Additionals:

Simultaneous alignment and clustering of peptide data using a Gibbs sampling approach.Bioinformatics. 2012 Oct 24. [Epub ahead of print]

9.00 - 9.30: Questions to yesterdays lectures and exercises
9.30 - 10.00: Data redundancy reduction algorithms (Hobohm1 and Hobohm2) [PDF]. [PPTX]. (MP4).
10.00 - 10.45: Optimization procedures - Gradient decent, Monte Carlo; Optimization procedures [PDF] [PPTX] (MP4); GD handout
10.45 - 11.00: Break
11.00 - 12.00: Gibbs sampling and Gibbs clustering; Gibbs sampling [PDF] . [PPTX] . (MP4).
12.00 - 13.00: Lunch
13.00 - 17.00: Hobohm data redundancy reduction algorithms; Answers to Hobohm programming exercise; Implementation of a Gibbs sampling algorithm for prediction of MHC class II binding; Answers

Wednesday, 11. June

Hidden Markov Models
Morten Nielsen

BACKGROUND TEXTS

9.00 - 9.30: Questions to yesterday lectures and exercise
9.30 - 12.00: Hidden Markov models; Viterbi decoding, Forward/Backward algorithm, Posterior decoding, Baum-Welsh learning; [PDF]. [PPTX] HMM (MP4); Viterbi Handout; Answers; Forward Handout; Answers
11.30 - 12.00: Profile Hidden Markov Models
12.00 - 13.00: Lunch
13.00 - 17.00: Implementation of Viterbi and posterior decoding.; Hidden Markov exercises; Answer to Hidden Markov exercises

Thursday, 12. June

Cross validation and training of data driven prediction methods. Stabilization matrix method (SMM)
Morten Nielsen

BACKGROUND TEXTS

9.00 - 9.30: Questions to yesterdays lectures and exercises
9.30 - 10.00: Cross validation and training of data driven prediction methods; [PDF] . [PPTX] . [MP4].
10.00 - 10.45: Stabilization matrix method (SMM) background; [PDF] . [PPTX] . [MP4].; SMM handout
10.45 - 11.00: Break
11.15 - 11.30: Description of potential projects and formation of groups; Project suggestions, and descriptions.; Document for project signup.
11.30 - 12.00: Quiz with questions capturing essential parts of course up to now Quiz. We will go over the answers to the quiz tomorrow morning.
12.00 - 13.00: Lunch
13.00 - 17.00: Implementing and evaluating SMM algorithms using cross-validation; Answers

Friday, 13. June

Artificial neural networks. Sequence encoding, feedforward and backpropagation algorithm
Morten Nielsen

BACKGROUND TEXTS

Immunological Bioinformatics. MIT Press. Chapter 4.

Background
Sequence encoding
Feed forward algorithm
Back-propagation and neural network training

9.00 - 9.30: Questions to yesterdays lectures and exercises
9.30 - 10.30: Artificial neural networks[PDF] . [PPTX] . Artificial neural networks Part 1 (MP4). Artificial neural networks Part 2 (MP4).; Handout; Answers
10.30 - 10.40: Break
10.40 - 12.00: Network training - backpropagation; Training of artificial neural networks [PDF] . [PPTX] . (MP4)..; Handout
12.00 - 13.00: Lunch
13.00 - 17.00: Artificial neural networks (Feedforward and Backpropagation); ANN answers

Monday, 16. June

An introduction to Deep neural network architectures
Morten Nielsen

BACKGROUND TEXTS

An introduction to deep learning on biological sequence data: examples and solutions

9.00 - 9.30: Questions to yesterdays lectures and exercises
9.30 - 10.00: Trick for ANN training [PDF].
10.00 - 10.30: NNAlign, alignment using ANN's [PDF]
10.30 - 12.00: Deep Learning using FFNN and NNAlign (PPTX) [PDF]
11.00 - 12.00: Exercise: Constructing and training Deep ANN methods (Joakim Noeddeskov Glifford && Jonas Birkelund Nilsson); NNdeep exercise
12.00 - 13.00: Lunch
13.00 -17.00: Exercise: Constructing and training Deep ANN methods (Joakim Noeddeskov Glifford && Jonas Birkelund Nilsson) cont.; NNdeep exercise; Answer Deep FFNN

Tuesday, 17. June

An introduction to Deep neural network architectures
Introduction to the project work
Morten Nielsen

BACKGROUND TEXTS

9.00 - 9.30: Questions to yesterdays lectures and exercises
9.30 - 10.00: Doing things in C - A few examples: Alignment and ANN training; Some code examples
10.00 - 12.00: Selection of projects, formation of project groups and start of project work Document for project signup.
11.00 - 17.00: Work on project (on your own)

Wednesday 18. - Tuesday 25. June. Project work

No lectures. Project work

Projects must be submitted (in PDF format) via campusnet Tuesday 24. of June 11.59 (just before lunch)

Wedensday, 25 - Thursday 26th June, Project evaluation and Exam

Each group has 15 minutes to present their project including 5 minutes for questions. Note, only the group will be present for the presentation of the individual projects. After the present, each member of the group is evaluated in a oral exam covering the complete course curriculum.

The exam will take place in the usual class room, building 210, room 042/048

Wednesday June 25th

8.00 - 8.50
Group 1	
Gibbs sampler approach to the prediction of SGBP binding
sites including pseudo counts and sequences weighting clustering
(Hobohm) techniques	
Natalia (s250668), Maria (s250352), Xiaopeng (s194408)

8.50 - 9.25
Group 2	
Comparation of FFNN and NNAlign for MHC binding prediction	
Marco (s243116), Martina (s243118)

9.25 - 10.10
Group 3	
Comparative study of PSSM, ANN, SMM for peptide MHC binding	
Nicoline (s203530), Mathilde (s215063), Alberte (s215067) and Kristine (s215098)

10.20 - 11.05
Group 4	
Compare PSSM, ANN, and SMM - predict MHC binding	
Annekatrine (s225074), Bunia (s215085), Emma (s215090)

11.05 - 11.50
Group 5	
Comparative study using ANN with different sequence encoding schemes	
Magnus (s204581), Oliver (s204692), Johanna (s204657), 

12.45 - 13.40
Group 6	
Comparative study of PSSM, ANN, SMM for peptide MHC binding	
Xavi (s243360), Luis (s243302), Miguel (s243284) and Natalia (s243547)

13.40 - 14.35
Group 7	
Comparative study of ANN with sparse, blosum and BERT encoding	
Astrid (s193254), Jui-Tse (s244401), Dinis (s212484), Sebastian (s233425)

14.35 - 14.50
Karolina Tudelska (s212846)

14.50 - 15.05
Saxe � Dali Wagner (s204559)

Thursday June 26th

9.00 - 9.55
Group 8	
Comparative study of ANN with different peptide encodings	
Pablo (s243357), Victor (s243634), Antonio (s243171), Elena (s243312)

9.55 - 10.40
Group 9	
HOBOHM-1 redundancy filtering of MHC class I binders in training data	
Tobias (s215105), Asbj�rn (s215045), and Henrik (s215065)

10.40 - 11.25
Group 10	
Implementation of a Baum-Welsh algorithm and/or Gibbs-sampling
method for training of an HMM	
Maximo (s247270), Tomas (s250441), Mikkel (s184240)

11.25 - 12.20
Group 11	
Comparative study of FFNN and NNAlign for peptide MHC binding	
Mei Lin (s194685), Magnus (s224188), Suru (s233406), Anne Sofie (s244028)

Go to

Main Course Page