Algorithms for Network-based Bioinformatics

The course will provide a brief overview on molecular biology, the advent of high throughput measurement techniques and large databases of biological knowledge. With this given background, the course will dive into the importance of various network approaches to model biological knowledge.

We will cover topics such as complex network models to characterise and analyse biological systems, approaches to infer network structures from given biological measurements, strategies of network enhancement through network integration, predictions based on network structures, and graph generation (molecular design).

Specifically, the course contains the following topics:

• Background on molecular data: systems biology, data-driven approach, high throughput measurements, sequencing, mass-spectrometry, immunoprecipitation, single cell measurements, molecular properties, molecular descriptors, databases.

• Molecular interactions: transcription factors, protein complexes, metabolic reactions, gene-protein-metabolite interactions, interaction models.

• Frameworks to work with complex systems: graphs with nodes/edges, directionality, layers, heterogeneous graphs, topological graph metrics, graph characteristics (erdos-renyi/scale-free/hierarchical networks), hypergraphs, topological hypergraph metrics, simplicial complexes, simplicial complex metrics

• Network inference and simulation: rate-law modelling of RNA regulation model, dynamic system/steady state, reduce, association networks, linear networks, design matrix, bayesian networks, boolean networks, regularisation, evaluation

• Network enhancement: heterogenous networks, reliability of networks, bias/variance, evaluation strategies, early/intermediate/late integration, machine learning approach, decision trees, classifier combining, probabilistic/dissimilarity/kernel-based representations, network fusion, random walks.

• Network-based predictions on the node-level: guilt-by-association, maximise coherency, Markov random field, clustering based on graph properties, Markov clustering, spectral clustering, Louvain method, active modules, validation.

• Network-based predictions on the graph level: graph neural networks, graph convolution, node message passing, graph attention, molecular property prediction, structure-based function prediction.

• Graph generation for molecular design: generative AI for molecular design, graph variational autoencoders, graph diffusion models

The final grade of the course consists of the following components:

• Individual Report (weighting 80%)

• Individual Presentation: presentation on a scientific paper (weighting 10%)

• Paper discussion: discussion points on paper (weighing 5%)

• General discussion: active role during discussions (weighting 5%)

Final grade calculation = 0.8 * Individual Report + 0.1 Individual Presentation + 0.05 Paper Discussion + 0.05 * General Discussion

A passing final grade for the course can only be earned when in both components at least a 5.0 is earned, and the weighted final grade is at least a 5.8, and the discussion point submission is a pass.

In case of an insufficient final result, repair options may exist in accordance with Article 17A, Times and number of examinations, sub 1, of the Teaching and Examination Regulations, for:

• Individual Report: Repair opportunity

• Individual Presentation: Repair opportunity

• Paper Discussion: Repair opportunity

• General Discussion: Repair opportunity

Disclaimer: information may change depending on unforeseen circumstances or measures (see: TER Art 2, sub 5).