Giovanni Misitano, Doctoral Student at University of Jyväskylä

Multi-objective optimization handles problems with multiple conflicting objectives with various tradeoffs: low cost leads to a high, negative environmental impact, for instance. Multi-objective optimization problems have many compromise solutions, which are not comparable on a pure mathematical basis. The aid of a domain expert, known as the decision maker, is employed. The decision maker can provide preference information, which can then be used to find the best possible solution among the compromise solutions. 

However, as decision-support tools, multi-objective optimization methods are often used to make real-life decisions with real-life consequences. If we do not fully understand how the tools we use operate, can we really justify the decision we make using them? I argue that many of the methods utilized in multi-objective optimization are pretty much just black-boxes to a decision maker. Can we address this problem?

Explainability has now been an established concept in the research of AI. It has spawned its own research field: explainable AI. Explainability has been used to address many of the problem encountered in AI, such as identifying biases in the data used and to better understand predictions made by black-box models. As a concept, however, explainability can be expanded to address any decision-support tools---not just AI.

I have been pioneering a new paradigm in decision-support tools: explainable multi-objective optimization. In my past works, I have taken inspiration from explainable AI, but in the future, we must developed novel and unique ways of incorporating explainability in multi-objective optimization to address its specific needs. As we move more and more towards a society governed by data and data-based solutions, it is important that we understand the tools we use and that the tools themselves are able to explain their behavior. Because only the best decisions can be also justified.

On June 24 the Leiden Centre for Computional Life Sciences organized a matchmaking event for researchers from the Faculty of Science. The goal of the event was to enhance crossover collaboration between different disciplines. At the event Dr. Doron Gollnast from the LIACS Project Office gave a presentation about the possibilities for funding interdisciplinary research within a European context.

At the Matchmaking Event, the following researchers presented their field of interest and future research plans. Click on each name to see a pdf of the slide presentation: 

- Gerard van Westen and Willem Jespers;

- Michael Emmerich; 

- Lu Cao; 

- Mitra Baratchi; 

- Alireza Mashaghi Tabari; 

- Paco Barona Gomez; 

- Evgeny Verbitsky (not presented at the meeting); 

- Daan Pelt (not presented at the meeting)

Jenke Scheen, PHD Student at The University of Edinburgh

Early-stage drug discovery campaigns pose major challenges to the pharmaceutical industry, mostly owing to large costs and high rate of failure of high-throughput screening of candidate molecules to the therapeutic target in question. Recent advances in computational chemistry and molecular simulation offer promising techniques that allow researchers to conduct these screenings virtually; especially Alchemical Free Energy (AFE) calculations have increasingly gained traction in solving the ligand-optimisation problem in both academic and corporate drug discovery [1]. Free Energy Perturbation (FEP) - where candidate molecules are transformed into each other virtually to compute the relative free energy of binding - has been one of the most tractable AFE techniques.

A critical step in FEP workflows is the generation of effective perturbation networks to ensure transformations are being simulated with sufficient phase-space overlap maximise prediction reliability. Currently, state-of-the-art softwares use primarily molecular similarity metrics to estimate optimal edges in networks. This method often still requires the user to review and tweak the presented perturbation network, hindering development toward a fully-automated FEP workflow. The current study uses a machine-learning (ML) approach to train models on many solvation FEP simulations to predict the reliability of a given perturbation a priori. Such models can replace similarity-based metrics to plan perturbation networks (figure 1).

This open-source project has resulted in several advancements that further progress the field of FEP. Using the data-driven method, practicioners can generate more reliable FEP networks, are able to transfer-learn to alternative FEP software reliabilities and fine-tune the model to the molecule series that is being investigated. The training domain used in this work has been made publicly available to allow researchers to create alternative novel models in the context of FEP.

Stefan Stemrau, Associate Professor at Leiden University

Slide presentation

Jan van Rijn, Assistant Professor in Machine Learning at LIACS

Modern AI solutions have a high dependency on large amounts of data. However, especially in life science applications, data is sometimes hard to acquire. The absence of this can lead to sub-optimal performance, which in turn limits the utility of the AI solution. In particular, deep neural networks can be very data-dependent. Data acquisition is an expensive part of the machine learning pipeline, which is in machine learning research projects often taken for granted.
However, for critical applications, access to labelled data can be the bottleneck. One way of addressing this data-dependency is by utilizing data from a similar source domain, and transferring the acquired knowledge to the target domain. The overarching field is called meta-learning.

In this talk, I will highlight several of the recent advances in meta-learning. I will briefly explain the assumptions that need to be met to apply it, and explain the basic methods MAML and Reptile. I will further explain some work that my group at LIACS does to better understand the underlying techniques. I will showcase some of the use-cases in life-sciences where we are successfully applied meta-learning to improve the utility of few data points.

Bas Goulooze, PhD Candidate at Leiden University

Children admitted to the intensive care can require large amounts of sedative drugs and painkillers. This puts them at a great risk for withdrawal symptoms once treatment with these drugs is stopped. We need more knowledge to avoid withdrawal in children, but available data is limited and many research questions remain unexplored in clinical studies due to ethical constraints. In this presentation, I will present how we leveraged available literature knowledge in combination with a novel modelling technique to analyse clinical withdrawal data in children. Using the model, we simulated ‘what if’ scenarios to generate new clinical hypotheses.

Please click for the slides of the talk (PDF).

Bert van Duijn, Professor of Plant electrophysiology

The Plant Biodynamics Laboratory (PBDL) favours multidisciplinary research. Depending on the problems under investigation, PBDL establishes collaborations between experts from for example molecular genetics, molecular and electro cell physiology, biochemistry, organic chemistry, analytical mathematics, mathematical statistics and computational sciences.

The plant hormone auxin (IAA) belongs to a complex system of chemical messengers which have a pivotal role in the coordination of growth and development between different parts of the plants, such as for example the shoot and root system.

We investigate the dynamics of long-distance polar auxin transport (PAT) by mathematical and computational analysis of experimental transport data. In particular, we investigate the functional role of the so-called PIN proteins which are assumed to be the polar distributed intracellular auxin-anion export carriers as postulated in the chemiosmotic theory of PAT. Our results already suggest that in spite of all the indirect evidence PINs are not solely or not at all responsible for PAT.

In addition we investigate the green algae Chara as model system for intra and inter cellular transport processes. Although it is known that Chara produces the natural auxin IAA, its transport and mechanism of action in Chara were unknown. However, over the past few years we were successful to be the first in demonstrating the presence of directional transport of IAA comparable to polar auxin transport (PAT) in terrestrial plants. Our program is now aimed at understanding the intracellular mechanism of this transport.

Marnix Medema, Wageningen University

Plants, fungi and bacteria produce a wealth of specialized metabolites, which are a great source for natural product drug discovery. Due to the accelerated accumulation of omics data, computational methods have become more and more important to identify these molecules and to assess their biological activities. Here, I will highlight the work performed in my research group on using these approaches to accelerate natural product discovery. Specifically, I will discuss the use of computational approaches to investigate biosynthetic diversity across large numbers of genomes, integrative genome/metabolome mining to link gene clusters to molecules, and computational approaches to predict biological activities of molecules based on omics data. 

André Leier, University of Alabama

Computational Approaches for the Development of Gene Therapies

Unfortunately, there is no video available for this talk.

Richard Allmendinger, The University of Manchester

Evaluating candidate solutions by conducting an experiment, e.g. a physical, biological or chemical experiment, can be expensive, time-consuming and resource intense. Drug development is a prime example where optimization relies on experiments. This talk will provide an overview of some non-standard challenges arising due to experiments, such as non-homogeneous per-objective evaluation times in a multi-objective problem, dynamic resource constraints, and non-static drug libraries in a drug discovery problem. Some existing techniques to cope with these challenges will be introduced, and, finally, promising areas of future work discussed. 

Nataša Jonoska, University of South Florida, USA

Unfortunately there is no video available for this webinar.

DNA rearrangement is a process found on both developmental and evolutionary scale. The process itself and the molecular shape at the time of the rearrangement can be modeled through 4-regular graphs.  These graph models are illustrated through the rearrangement processes in a well studied ciliate species Oxytricha trifallax where DNA recombination is observed on a massive scale.  Our studies show gene segments that recombine during DNA rearrangement processes may be organized on the chromosome in a variety of ways. They can overlap, interleave or one may be a subsegment of another. We use colored directed graphs to represent contigs containing rearranged segments where edges represent recombining segment organization. Using graph properties we associate a point in a higher dimensional Euclidean space to each graph such that cluster formations and analysis can be performed with various methods. The analysis shows some emerging graph structures indicating that segments of a single gene can interleave, or even contain,  all of the segments from several other genes in between its segments.

Joost Batenburg, Professor of Imaging and Visualisation at LIACS

In this talk I will go over various types of tomographic imaging (CT scanning, electron tomography, optical tomography) that are used for biomedical and life science research. A key step in each of these imaging modalities is the tomographic reconstruction step, where the data acquired by the imaging instrument is turned into a 3D representation of the specimen. While this operation is usually hidden inside the instrument’s software, it can dramatically influence the resulting image quality, imaging artefacts, and as a result the ability to observe key internal processes with sufficient accuracy. 
I will discuss various recent breakthroughs in algorithms and machine learning models that enable to compute high quality images from severely undersampled data, and to do so in real-time, leading to an interactive computational imaging instrument.

Jeroen Codée, Professor of Organic Chemistry, LIC

Unfortunately there is no video available for this webinar.

Chemically synthesized carbohydrates (glycans/oligosaccharides) are indispensable in biochemical and medical research. Their synthesis is challenging because of the highly reactive intermediates that play an important role in connecting different carbohydrate building blocks. We have developed computational strategies to understand how the structure of these fleeting species governs their reactivity. Rather than studying a single example, we have mapped the reactivity of a large family of related species to establish structure-reactivity rules. The body of systematic experimental data that we are currently gathering may be used in the future to computationally predict the outcome of glycosylation reactions eventually enabling the in silico design of synthetic routes to fast-forward oligosaccharide synthesis and boost glycobiological and glycomedical research. 

Erik van den Akker & Boudewijn Lelieveldt @ LUMC

Unfortunately there is no video available for this past event.

Lu Cao, Assistant Professor at LIACS

Unfortunately there is no video available for this past event.

Barbara Scalvini, PhD candidate at LACDR

Unfortunately there is no video available for this past event.

Oliver Kramer, Professor at University of Oldenburg

Unfortunately there is no video available for this past event.

On the 14th of July CCLS held a Kahoot pub quiz on trivia, films, music, life sciences, informatics and more.

Unfortunately there is no video available for this past event.

Agnieszka Wegrzyn, PostDoc Researcher at Leiden Academic Centre for Drug Research, Leiden University

Unfortunately there is no video available for this past event.

Ariane Briegel, Professor of Ultrastructural Biology at Leiden University

Most motile bacteria contain a highly sensitive and adaptable sensory system composed of clusters of chemoreceptors. This chemosensory system is used to detect changes in nutrient concentrations, allows the cells to navigate towards preferential environments and is also involved in host infection by some pathogenic bacteria. While it is one of the best understood signaling systems to date, unraveling structure and function of the bacterial chemotaxis system remains challenging. High-resolution analysis using methods are inherently limited to structural fragments and rely on specimens taken out of their natural environment. Thus, they lack the larger context of the native system. Here, we use cryo-electron tomography (cryoET) to study the three-dimensional architecture of the bacterial chemoreceptor arrays. We identified species-specific characteristics of this system of selected pathogens, such as in Treponema denticola (periodontal disease) and Vibrio cholerae (cholera). However, not all bacteria are motile, but can still benefit from chemotactic behavior: we have gained insight how non-motile bacteria can benefit from chemotactic microbes that share the same environment.

Unfortunately there is no video available for this past event.

Programme

Ahmed Mahfouz 
Untangling Single Cell Data; Interaction, Identification and Intergration

Jelle Goeman
Double Dipping Done Right; Interactive, Statistical Inference in Neuroimaging

Patrick Echtenbruck
Innovage

Drinks & Nibbles at Foobar

Programme

Update on CCLS

Break out in Groups and Logo Design

Pitches by Groups and Discussion

Drinks & Nibbles at Foobar

Programme

Vivi Rottschäfer
Towards a 3D distribution Model of Drugs in the Brain 

Sander Hill
Computational Issues in Mathematical (Plant) Biology

Drinks & Nibbles at Foobar

Programme

15.00 - 15.10   Welcome with coffee/tea

15.10 - 15.30   Coen van Hasselt - Quantitative modeling in translational and clinical pharmacology (LACDR)

15.30 -15.50    Sirin Yunucu - Translational systems pharmacology modeling for immune-oncology (LACDR) 

15.50 - 16.30   Discussion

16.30 - 17.30   Drinks and Nibbles in LAB 071

Charles Hoyt, Fraunhofer Institute for Algorithms and Scientific Computing in Sankt Augustin, Germany: 

Drug discovery campaigns are often successful because of a coherence between the identification of key proteins in a disease and their subsequent targeting for modulation. Network representation learning will be presented as a schema-free framework for integrating and jointly modeling several biological relationships from heterogeneous data types like chemical activities, chemical similarity, drugs’ side effects, drugs’ indications, and disease-associated proteins. It will be shown how link prediction tasks in an integrative network are isomorphic to several key tasks in drug discovery such as proteochemometrics, target prioritization, side effect prediction, and drug repositioning. Finally, it will be shown how the underlying network can be directly used to identify and rank potential mechanisms of action for drug repositioning candidates.

Programme

16.00 - 16.10 Welcome with Coffee and Tea

16.10 - 16.35 Willem Jespers - Molecular Dynamics and Free Energy Calculation Approaches: What can it do for you

16.35 - 17.05 Alexandre Goultiaev - Structural Bioinformatics of Virus RNA's

17.05 - 17.15 Michael Emmerich - NWO Klein Proposal

17.15 - 17.30 Discussion and Announcements

17.30 Drinks and Nibbles at the Science Bar

Lunch at the Faculty Club

Julia Handl, Manchester University - Data-driven Optimization: Using machine learning principles to focus the search in evolutionary optimization

Workshop in the Hortus Botanics: Herbs and Spices or Tour of the Botanical Gardens

Lindsey Burggraaff LACDR - Challenges in Computational Drug Discovery - Bridging Computer Science and Medicianal Chemistry

Irene Martorelli LIACS - Integration and Computational A