This Week's Discoveries | 10 December 2019
- Daniel Rozen
- Mehrnoosh Sadrzadeh
- Tuesday 10 December 2019
Niels Bohrweg 2
2333 CA Leiden
- De Sitterzaal
Division of labour in multicellular bacteria.
Daniel Rozen (IBL)
Daniel is an Associate Professor at the IBL at Leiden University where he has worked for the last 7 years. His research focuses on the evolution and ecology of microbial populations, with a particular emphasis on the evolution of antibiotic resistance and the ecological roles of antibiotic production in nature.
One of the hallmark behaviours of social groups is the division of labour, where different group members become specialized to carry out complementary tasks. By dividing labour, cooperative groups of individuals increase their efficiency, thereby raising group fitness even if these specialized behaviours reduce the fitness of individual group members. We have recently discovered that antibiotic production in colonies of the multicellular bacterium Streptomyces coelicolor is coordinated by a division of labour. We've shown that S. coelicolor colonies are genetically heterogeneous due to massive amplifications and deletions to the chromosome. Cells with gross chromosomal changes produce an increased diversity of secondary metabolites and secrete significantly more antibiotics; however, these changes come at the cost of dramatically reduced individual fitness, providing direct evidence for a trade-off between secondary metabolite production and fitness. Finally, we show that colonies containing mixtures of mutant strains and their parents produce significantly more antibiotics, while colony-wide spore production remains unchanged. Our work demonstrates that by generating mutants that are specialised to hyper-produce antibiotics, streptomycetes reduce the colony-wide fitness costs of secreted secondary metabolites while maximizing the yield and diversity of these products.
Second lecture, Lorentz Center Highlight
Understanding natural language: logic or statistics, which one is more important?
Mehrnoosh Sadrzadeh(University College London )
Mehrnoosh Sadrzadeh is a Reader in Computer Science at University College London. Her research is focused on developing high-level logical and mathematical models for computer systems, learning their parameters from data, e.g. via machine learning, automating the reasoning using proof-theoretic and algebraic tools, and applying the results to mainstream and industrial tasks. Her current interest is Natural Language Processing (NLP), previously she led work on Multi-Agent Systems (MAS). Mehrnoosh is one of the organizers of the workshop “Logic and Structure in Computer Science and Beyond (LSCSB)” that is being held in the Lorentz Center from 9 December through 13 December 2019.
The AI fields of computational linguistics and natural language processing have seen the rise of two very different challenges in the past two centuries. In the 20th century, the main concern was how logical structures can formalise grammatical rules. With the ever-growing amount of easily accessible data, the 21st century has favoured statistical modelling of linguistic units as points in highly dimensional spaces. But what is more important when it comes to modelling natural language, logic or statistics?