Personal tools
You are here: Home People Loewe, Laurence Research

Research interests of Laurence Loewe

My research centres on detailed simulation models that integrate biological knowledge in order to understand more about evolution. To do this I combine current systems biology with evolutionary genetics in an approach that I call evolutionary systems biology. I use this approach to investigate the genetic basis for adaptive evolution and the evolutionary consequences of slightly deleterious mutations. These general topics are important for understanding the evolution of antibiotics resistance in bacteria and many other practical problems.

The collection of research questions below inspired me to get into developing the programming language Evolvix in order to reduce the work needed for answering them.

While developing Evolvix, I have continued my work in population genetics (estimate measures of diversity to infer selective sweeps and background selection) and molecular systems biology (predict circadian clocks in Drosophila), even if not yet published (working on it).

The combination of all these topics has been terrific: working on a non-trivial biological research problem while simultaneously designing a compiler for a general programming language for biology has helped me to see many innovative approaches for how Evolvix could simplify the life of computing biologists. Doing both may have appeared to slow me down, but in reality the combination provided me with unique insights that are usually lost on compiler builders (as they don't do biology) and on computing biologists (as they are often unaware of the enormous power provided by compilers).

Once Evolvix is complete enough for use in most of my research (not just for toy models as used in Ehlert & Loewe (2014), I will pick up the trail of the questions below. To meet this goal I have decided to extend the declarative approach of Evolvix (best for many models) and add general programming capabilities (for saying what to do with the models). The first goal is to make Evolvix Turing complete, then add more general structured programming capabilities and leave other paradigms for later (while taking care not to step on the territory reserved for object-oriented, functional, or other programming paradigms of use in biology).  

Currently Evolvix leads a hybrid life: it has a working public prototype that is freely available and already quite useful in research (see Ehlert & Loewe, 2014) and teaching (used since 2013 in the Evolutionary Systems Biology Course at UW-Madison). It is intended to become open source as soon as reasonably possible, but it is not there yet (we need to clear out much of the old backend C++ code which is unnecessarily cumbersome to explain and use, hence not worth anybody's time; once the new code is in place, we will publish as soon as we can). There is a terrific design and a very rough roadmap that I am really excited about, but it is not written down in a form that makes sense to share without further explanations.

If you are interested in supporting the development of the first general programming language designed by biologists for biologists, then please feel free to contact me.

We need

  • excellent programmers, language designers, compiler builders, data modeling experts, C++, JavaScript, Python etc coders; if you are outstanding  and/or have industry experience, please let us know, we have jobs coming up.
  • domain experts like biologists, chemists, physicists, engineers, statisticians and mathematicians to help us refine our modeling approaches. 
  • reviewers at many levels, from experts who help us to make sure what we do and say is accurate, to beginners who help us to make sure that the new features are easy to understand and use.
  • testers at many levels (alpha, beta, etc) who help us to ensure that Evolvix actually works out of the box for as many people as possible (we already have an automated build system for some basic tests). 
  • editors and writers who help us write an excellent manual that actually answers the questions of Evolvix users, along with various examples for how Evolvix can be used in various disciplines.

To coordinate that many contributions, we have been developing a system that is independent from the usual development tools like Git, Jira, and Confluence (which we all use). We are now in the process of setting up this collaborative editing system, mostly to facilitate our own development process and move the Evolvix design from scribbled paper to properly formatted documents.  Contributing will become *much* easier, once this system of "Evolvix DesignDocs" is up and running in a publishable form. We are getting close, but are not there yet.

Let us know if you think you can help in some capacity and we will let you know if we have something, where you could contribute now, soon or maybe later.

If you think developing something like Evolvix is a terrific idea and would like to support our project financially, please let us know.



General biological research interests

My research focuses on understanding the molecular, genetic and ecological basis for adaptive evolution and extinction in natural populations and its implications. This means that I am interested not only in understanding the long-term evolution of species, but also the multitude of factors that play crucial roles in the success of a species. On one end of the scale this includes spatial structure, local recombination and mutation rates, as well as life-history details and environmental changes. On the other end of the scale I am interested in very detailed intracellular molecular systems biological simulations that will hopefully contribute towards a better understanding of the frequency with which certain fitness changes occur. I am interested in a wide range of species, including bacteria that can evolve antibiotics resistance.

Since the biological reality of natural populations is exceedingly complex, it is not possible to construct mathematically rigorous completely realistic 'true' models of evolution. Historically, much progress in evolutionary theory has derived from simple models that capture the essence of a few important features of evolution and allow a full understanding of this reduced universe. I am interested in bridging the gap between simple analytically understandable mathematical models and biological reality by building rigorous simulation models to understand the following big topics:

  • Origin and extinction of species. I want to understand the various processes that can lead to extinction, prevent extinctions and lead to the origin of new species. This is not only important for designing effective strategies for long-term conservation of endangered species, but also for understanding what makes potentially harmful bacterial strains survive or go extinct.
  • How does evolution shape sequence patterns in genomes? This holds vital clues for understanding our genomes and is important for testing our understanding of evolution. A proper understanding of evolution is key for devising strategies to solve practical problems.
  • Evolutionary consequences of human interferences with natural processes. We change our environment in more ways than most people realise. If we want to preserve this planet on the long term, we need to understand the evolutionary impact of our behaviour, from spreading mutagens over habitat destruction to the release of genetically modified organisms. How we use antibiotics in medicine and agriculture has enormous consequences for the evolution of anibiotics resistance in bacteria, which has direct medical consequences.
  • Evolution of novel features. Such features can lead to multiple resistance and pathogenicity in microbes that can cause widespread dangerous infections. A thorough understanding of the evolution of these features is pivotal to finding the best strategies to reduce the corresponding threat. The generality of the theory of evolution often allows us to apply findings in one species to other cases as well.  
  • The distribution of mutational effects and epistasis. Many DNA changes have some effect, even if it is very small. It is difficult to estimate how many DNA changes have a particular effect and how these effects change if many mutations accumulate in a particular system. To address these questions I have developed a framework for evolutionary systems biology, which uses simulations of the biochemical reactions of a particular molecular system to infer those fitness effects that are too small to be measured in the laboratory based on calibrations from mutations with larger effects. Understanding these small effects is crucial for understanding long-term evolution in any system from microbes to man.


Evolutionary systems biology research interests

Research in evolutionary genetics has shown that the fitness effects of DNA changes determine the long-term outcome of many evolutionary processes. I have developed and used a population genetical approach to estimate the distribution of deleterious mutational effects for the whole genome from a sample of DNA sequence diversity. This approach can give a good bird's eye perspective overview, but it is always better to confirm results using an independent method. I developed a new approach that builds on simulations of intracellular molecular systems to compute fitness correlates that can inform us about fitness changes (for more on this evolutionary systems biology approach see the full description). This fuels my interest in the following questions.

  • Fitness correlates. Is it possible to define fitness correlates that allow us to predict the selection coefficients of a particular molecular system from the output of a systems biological simulation? Initial analyses provide some hope and I want to test this concept on the following three systems.
  • Circadian clocks. These molecular systems are virtually everywhere and help organisms to distinguish day and night. Quite a bit is known about their molecular mechanisms, so there is some hope that corresponding systems biological models are somewhat realistic. I am interested in exploring the parameter space of working clocks. For example, how difficult is it to improve or destroy a clock? An initial study has alrady led to interesting results and I plan to analyse more complicated clocks too.
  • Signal transduction pathways. Using the cholesterol pathway as an example, I am conducting systems biology simulations that help to understand what molecular changes can affect this important signalling system.
  • RNA metabolism. This system's central role to genome function and gene regulation implies a correspondingly large potential for fitness relevant effects. I plan to build a mechanistic model that helps to explore the various potential effects of molecular changes in detailed models of RNA metabolism on fitness.
  • Methods of simulation. Randomness and noise are prevalent features of molecular systems where frequently only a few components play vital roles. This seriously limits the predictive power of ordinary differential equations and thus I employ stochastic simulations that track all the molecule counts over a time course for a particular run of the system. These time courses are then automatically analysed to compute the most interesting results. Since these simulations are very computing intensive, I work on providing the possibility to use evolution@home to compute the results.


Evolutionary genetics research interests

Before my work in current systems biology I have worked on the following questions that still interest me:

  • Extinctions caused by Muller's ratchet. Muller's ratchet is a population genetical process that can lead to deleterious mutation accumulation in a population. It has long been known to be capable of causing extinctions of species under various circumstances, but quantification of such circumstances have remained scarce. Using evolution@home, I built the largest existing database of Muller's ratchet simulation results. The accumulated results of many decades of CPU-time help to bring more precision to tests of the hypothesis that various species can or cannot be driven to extinction by Muller's ratchet.
  • The strength of selection. While theory frequently uses selection coefficients to measure how strong selection will favour a particular genotype, too little is known on what these numbers actually are in nature. I am helping to put some numbers to the theory, as this is crucial for running meaningful simulations.
  • The distribution of mutational effects (DME). It is well known that mutations can have widely varying effects on fitness. A DME quantifies the probability that a new mutation will have a particular strength of selection. I helped develop a new method for estimating DMEs that was first tested in two species of fruitfly (Drosophila miranda and D. pseudoobscura). I am interested in expanding application and sophistication of this and similar methods.
  • Effects of deleterious and adaptive mutations on patterns in genomes. We know that most mutations of any effect have harmful effects. I am interested in using theory and simulations to predict how observable patterns in genomes are shaped by these recurrent deleterious mutations. Theory also predicts that advantageous mutations leave clear signatures in genomes when they sweep to fixation. I am interested in helping to disentangle these from the various similar looking effects (e.g. of changes in population size). I hope this will improve our estimates of the frequency of advantageous mutations. To address these questions I am planing a massive simulation project that models many different mutations of different effects in different individuals with varying degrees of recombination and mutation. This will be implemented as a new simulator of evolution@home.


Application to antibiotics resistance evolution

One of my long-term goals is to to combine my interests in all these diverse topics for addressing a question of immense practical importance: How can we minimise the evolution of antibiotics resistance evolution? This question is one of the big medical challenges of today, as the evolution of resistant 'superbugs' threatens to reverse one of the most important medical advances of the 20th century. I plan to develop a series of increasingly realistic simulations in order to investigate the effectiveness of various antibiotics usage policies that are designed to minimise resistance evolution in bacteria like E. coli. Since the realism of simulations is often limited by the computational complexity that can be handled by the underlying computing system, I am interested in developing approaches that facilitate analysing such models of evolution.


Computing research interests

All following computational research interests have now been merged/expanded into my work on Evolvix. In due course I will re-write these for the Evolvix website, but for now, I'll leave up the list of topics from before I started Evolvix:

Most of my biological questions are not easy to answer, as they involve many interacting entities on many different levels that can be very difficult to capture in analytical mathematical models. I use various mathematical tools, but when normal tools can no longer handle the complexity, I resolve to stochastic simulations run by evolution@home to do the heavy lifting. In 2001 I started evolution@home exactly for this purpose. It is the first global computing system for evolutionary biology and I am continually developing it, slowly, but steadily. This fuels my interest in quite a number of computer science related topics like

  • Grid/Cloud computing. Global computing is just a special case of grid computing. It uses the Internet to distribute tasks to computers from the general public that volunteer for doing some work.
  • Databases. Since computation of results can be extremely costly, it makes sense to store results for fast analyses. I am interested in efficient large-scale datawarehouse organization to allow scalable handling of complex collections of results.
  • Visualization. Complex results are often better understood if visualized appropriately. I am interested especially in visualization of multi-dimensional results.
  • Statistical methods and machine learning. Each simulation produces much more data than can possibly be stored on any central server, no matter how powerful. Thus I design my simulators in such a way that they extract the key results of a simulation for central collection. I am interested in making this automated extraction of observations as powerful as possible to make the best use of the publicly contributed simulation time. This involves automating statistical analyses. I am also interested in how the final results collections can be analysed. This includes an interest in parameter estimation and approaches for Approximate Bayesian Computation.
  • Algorithm design. Complex simulations can benefit tremendously from clever algorithm design that can cut required computing times by a large margin.
  • Security. Any global computing network has to deal with security issues. So it is only natural that I am interested relevant issues as well.

To glue all these topics together with the analysis of actual sequence data I have a long-standing interest in bioinformatics.



Document Actions