DDLS Annual Conference

Title: Deep Learning for Microscopy

Affiliation: EMBL, Heidelberg

Bio: Anna Kreshuk is a group leader and a Senior Scientist in the Cell Biology and Biophysics Unit at the European Molecular Biology Laboratory (EMBL) in Heidelberg, Germany. She received a PhD in Computer Science from Heidelberg University in 2012. Following this, she worked as a PostDoc in the Heidelberg Collaboratory for Image Processing (HCI). She joined EMBL in July 2018, where her research group develops novel computer vision methods for the analysis of microscopy images. Beyond method development, she is also committed to the democratisation of machine learning for the life science community, contributing open source tools such as ilastik and PlantSeg. She is an ELLIS scholar and a scientific co-coordinator of the Horizon Europe project AI4Life.

Abstract:

Title: The archaeal evolutionary origins of the eukaryotic cell

Affiliation: Utrecht University, the Netherlands

Bio: Daniel Tamarit is an Assistant professor at the Theoretical Biology and Bioinformatics (TBB) group, specialized in comparative and evolutionary genomics. He obtained his PhD at Uppsala University by studying the evolution of host-associated bacteria, and performed postdoctoral research at Wageningen University on major evolutionary transitions such as the origin of eukaryotes. Currently, his team uses evolutionary genomics to investigate fundamental questions in archaeal and bacterial biology, such as (i) what are the principles governing bacterial and archaeal genome architecture, (ii) what can the genomes of novel microbial lineages tell us about basic evolutionary processes, and (iii) what is the shape of the tree of life. 

Abstract: The origin of eukaryotic cells was one of the most impactful transitions in evolution, setting the basis for the emergence of multicellular organisms such as animals, plants and fungi. Recent advances in genome-resolved metagenomics and phylogenomics have placed Asgard archaea at the heart of this transition, revealing that eukaryotes originated from within this group. Expanded genomic sampling and robust evolutionary modeling identify a lineage of Heimdallarchaeia, the Hodarchaeales, as the closest known relatives of eukaryotes. Reconstructions of ancestral gene repertoires, structure-based homology searches, and targeted experiments suggest that the archaeal progenitor of eukaryotes already possessed an enriched set of genes involved in cytoskeletal dynamics, membrane trafficking, and other functions that contribute to cellular complexity. This talk will showcase how integrating computational approaches is transforming our understanding of early cellular evolution and revealing the deep archaeal roots of eukaryotic life.

Title: Discovering patterns in the codon sequences of proteins

Affiliation: Center for Molecular Protein Science, Lund University

Bio: Prof. André has a PhD in Biophysical Chemistry from Lund University. He joined the University of Washington for postdoctoral work before returning to Lund University to start his research group in 2009. He studies protein self-assembly to understand the mechanism behind the formation of large assemblies using experimental and computational methods. Large assemblies are also designed using computational protein design methods and artificial intelligence. His group also works on the design of conformational changes in proteins. Another direction is the study of codon adaptation in organisms and its effect on co-translational folding. Questions related to how protein structure has evolved are also addressed in research by the group using computational and experimental methods.

Abstract: The genetic code defines how amino acid sequences are encoded in mRNA. Due to redundancy, this encoding is not unique, with up to six different codons specifying the same amino acid. Historically, synonymous mutations, changes between codons for the same amino acid, have been referred to as “silent.” However, it is now clear that codon choice can have significant functional consequences.

We investigated the effect of synonymous codons on protein folding efficiency in E. coli using a high-throughput experimental assay. Our results reveal a link between protein topology and the proportion of nascent proteins that emerge from the ribosome correctly folded, showing that even single synonymous substitutions can strongly influence co-translational protein folding.

Complementary bioinformatic analyses establish a relationship between codon usage and protein secondary structure. Furthermore, by training a large language model, we developed a predictive framework that identifies positions of rare codons within coding sequences, uncovers patterns explaining codon choice along the sequence, and highlights mRNA stability as an important determinant of codon usage.

Title: Restless bandits and self-driving microscopes

Affiliation: KTH Royal Institute of Technology

Bio: Joakim Jaldén has a M.Sc. in electrical engineering and a Ph.D. in telecommunications from the KTH. Since his employment as a tenure-track faculty at KTH in 2009, he has increasingly devoted the group’s research efforts towards data analysis challenges in life science applications. Highlights include: The development of data association algorithms for single-cell tracking in time-lapse microscopy, in collaboration with the Blau lab at Stanford; Mathematical models and algorithms for cytokine movement and antibody interaction in immunoassay, which laid the foundation for patented technology behind the Mabtech AB analysis instruments for ELISpot and FluoroSpot, and most recently, multiplex assays; Algorithms and AI for long-read DNA basecalling from nanopore data, in collaboration with SciLifeLab. He has been a full professor at KTH since 2017, where he is now also deeply engaged in education.

Abstract: The talk will focus on one result stemming from a first round of joint WASP-DDLS projects. The proposal envisioned the use of reinforcement learning as a tool to capture images of rare cell states in the context of automated microscopy. It will cover, at a conceptual level, the mathematical modeling decisions made to arrive at principled automation rules regarding what field of view to image and for how long, ending with empirical illustrations of the practically realizable gains.

Title: The evolutionary history of a commensal species: lessons from adapting to the Anthropocene

Affiliation: Centre for Ecological and Evolutionary Synthesis, University of Oslo

Bio: Mark is an evolutionary biologist interested in understanding speciation and adaptation using a wide range of ecological and genomic techinques. He completed his PhD on local adaptation and speciation in three-spined sticklebacks at Queen’s University Belfast. Following that he was a JSPS Postdoctoral Fellow at the National Institute of Genetics, Japan, where he worked on the genomics of speciation in marine sticklebacks. He continued his work on the genomics of speciation and hybridisation, this time on Passer sparrows, at the University of Oslo as a Marie Curie fellow. It was here that his interest in understanding how human activity has shaped the evolution of other species began. He also has a strong interest in teaching genomics and bioinformatics to those beginning in the field. He has written several well-known resources and is co-author on the textbook Evolutionary Genetics: Concepts, Analysis and Practice. He is now an Associate Professor at the University of Oslo and his primary research focus is on using Passer sparrows as a model system for understanding how human activity might have driven adaptation, hybridisation and even potentially speciation in these charismatic little birds.

Abstract: Human activity has altered the evolutionary trajectories and ecological circumstances of nearly every species on Earth. Commensal species are an extreme example to adaptation to anthropogenic niches, benefitting from a close association with human activity. What can the evolution of a commensal life history tell us about how species adapt to environmental change driven by humans?
House sparrows (Passer domesticus) are a hugely successful human-commensal species, occurring on nearly every continent, largely due to human introductions. Despite its close association human society, surprisingly little is known about the evolutionary history of this species, how it has evolved to fit a tight anthropogenic niche and the evolutionary consequences of its rapid global spread. Our work has focused on addressing this.With resequencing data from over 1200 Passer sparrows, we first reconstructed the evolutionary history of the house sparrow. Our results suggest the species likely adapted to a human niche, potentially more than once, in Central Asia during the last 10,000 years. We used phenotypic data including micro-CT 3D scanning to compare skull morphology between human commensal and wild sparrow populations. We identified clear divergence in skull morphology, beak shape and an increased brain size in the commensal house sparrow populations. A major signature selective sweep in the house sparrow genome encompasses two major candidate genes; COL11A1 – which regulates craniofacial and skull development and AMY2A which is linked to adaptation to high-starch diets in humans and dogs. Functional analysis of COL11A1 confirms its role in skull development and we show evidence of copy number variation at AMY2A.
Our next aim was to investigate the more recent global spread of the species. Native to Eurasia, the house sparrow was introduced to the Americas, Australasia and Sub-Saharan Africa in the 19th and 20th centuries. Using resequenced genome data from across this range, we have reconstructed the population genomic signatures of these introductions. Our findings point to introductions from two divergent lineages of house sparrow to different parts of the world as well as the rapid evolution of genetic differentiation among introduced populations. We further investigated adaptation to climatic variation across the introduced ranges of the species in North America and Australia. We identify several candidate genes linked to growth, thermoregulation and thermal tolerance, suggesting introduced populations have adapted to new environmental conditions in less than 150 generations.
Our work identifies phenotypes and genes involved in rapid adaptation in parallel to anthropogenic change across the native and introduced range of the house sparrow. With a global distribution spanning a wide range of climatic conditions and a close association with human society, we suggest the house sparrow is an ideal model species for understanding adaptation in the Anthropocene.

Title: T cell receptor diversity and immune health as a prognostic cancer biomarker

Affiliation: Department of Clinical Medicine, Aarhus University, Denmark

Bio: Nicolai Birkbak is Professor of Cancer Evolution and Bioinformatics at Aarhus University and Aarhus University Hospital, Denmark, where he leads an independent research group. His research focuses on the bioinformatic analysis of high-dimensional data to understand how the immune system interacts with cancer. By integrating genomic, transcriptomic, single-cell, and spatial data with medical imaging and liquid biopsy analyses, including circulating tumor DNA and T-cell receptor sequencing, his group develops holistic biomarkers that capture both tumor biology and host immune responses.

Nicolai trained in bioinformatics at the Technical University of Denmark and completed postdoctoral research at the Dana-Farber Cancer Institute and the Francis Crick Institute. At the Crick, he contributed to the TRACERx lung cancer evolution project, advancing understanding of how tumors evolve and evade immune surveillance. His work has also helped identify biomarkers of sensitivity to PARP inhibitors and elucidate mechanisms linking tumor evolution, immune modulation, and metastatic progression.

The overarching aim of his research is to translate these insights into precision medicine – enabling earlier detection, improved prediction of treatment response, and ultimately better outcomes for cancer patients.

Abstract: The adaptive immune system not only protects us against pathogens but also plays a crucial role in defending against cancer. Among its key components, T cells serve as central effectors in the endogenous anti-cancer response. Yet, the clinical significance of their quantity, diversity, and dynamics remains insufficiently understood.

We investigated the prognostic value of the T cell receptor (TCR) repertoire in patients with bladder cancer. In advanced-stage disease, we found that low pre-treatment peripheral TCR diversity was associated with poorer overall survival, particularly when combined with low circulating T cell fractions. These low-diversity repertoires were dominated by hyperexpanded clones that persisted throughout treatment and were frequently directed against chronic viral infections such as cytomegalovirus. Longitudinal analyses further revealed treatment-associated declines in TCR diversity, suggesting adverse effects on systemic immune health.

Together, these findings highlight that immune health biomarkers such as TCR diversity may provide new opportunities for precision medicine approaches aimed at improving cancer treatment outcomes.

Title: From tumor maps to new therapies for pediatric brain cancer

Affiliation: Department of Immunology, Genetics and Pathology; Uppsala University

Bio: Veronica studied Biology at Simon Bolivar University in Venezuela, and in 2013 moved to Sweden to pursue a PhD at Uppsala University. During this time, she worked under the mentorship of Prof. Tobias Sjöblom and studied how genomic losses in cancer could be exploited for therapy. In 2019, she moved to Boston (USA) to pursue a postdoc in Dr. Rameen Beroukhim’s lab at Dana-Farber Cancer Institute, affiliated with Harvard Medical School and the Broad Institute of MIT and Harvard. Here, Veronica focused on studying mechanisms of response and resistance associated with p53 reactivation in malignant brain tumors. She additionally continued her work on aneuploidy, describing a new class of therapeutic targets as ‘toxic genes’ – genes whose high-levels of expression are detrimental to cancer cell fitness. In 2025, Veronica became a Group Leader in Neuro-Oncology at the Department of Immunology, Genetics and Pathology at Uppsala University.  Her lab investigates the genetic and adaptive mechanisms driving brain tumor evolution and therapy resistance, with the goal of identifying new treatment targets.

Abstract: In Sweden, cancer remains the leading cause of disease-related death among children, with brain tumors representing the most lethal subtype due to their limited treatment options and high relapse rates. One rare and aggressive subtype is atypical teratoid/rhabdoid tumors (AT/RT), an embryonal brain tumor which affects infants of under 2 years of age. The low incidence of AT/RT has hindered the establishment of a universal standard of care. Because these tumors primarily affect infants under two years of age, treatment is largely restricted to intensive chemotherapy regimens designed to avoid the harmful effects of radiation therapy. However, these high-dose regimens often yield only partial responses, underscoring the urgent need to elucidate mechanisms of chemoresistance and to identify novel therapeutic strategies. We aim to uncover the transcriptional cell states and genetic vulnerabilities associated with chemotherapy-persistent AT/RT, thereby revealing new targets that can enhance the efficacy of frontline treatments. To this end, we first employ cellular barcoding to track cancer cell populations at single-cell resolution before and after exposure to individual chemotherapeutic agents used in clinical AT/RT protocols. This approach enables comparison of treatment regimens and identification of overlapping mechanisms of resistance. To further dissect these mechanisms, we leverage CRISPR sensitizer screens to define targetable genes that potentiate chemotherapy response when depleted. Through our integrated transcriptomic and functional genomic analyses, we ultimately aim to uncover novel therapeutic targets in chemotherapy-persistent AT/RT cells and accelerate the development of more effective treatment strategies.

Titel: Mapping infection at proteome scale: Toward mechanistic design of phage-based antimicrobials

Affiliation: Karolinska Institutet

Abstract: The rise of antimicrobial resistance (AMR) is outpacing the discovery of new antibiotics, creating an urgent need for alternative strategies. Bacteriophages (phages) offer a promising route to targeted antimicrobials, but their rational use is hindered by limited understanding of how they reprogram bacterial cells during infection. Our work addresses this by systematically mapping host–phage protein–protein interactions (PPIs) to uncover how viral proteins hijack essential bacterial complexes and redirect cellular physiology. Using next-generation interaction proteomics based on co-fractionation mass spectrometry (SEC-MS), we generated the first temporally resolved interactome of mycobacteriophage-infected cells. This approach captures native phage and host complexes without genetic tagging, enabling functional inference of phage proteins directly from their physical context. We identify hundreds of infection-specific assemblies, including viral inhibitors of essential bacterial machineries such as RNA polymerase and ribosomes. Structural modeling and comparative network analysis reveal conserved interface motifs and divergent strategies of complex remodeling across phage families. Integration of these data with CRISPRi perturbation and antibiotic susceptibility profiling provides a quantitative framework to predict infection outcomes and phage engineering.

Title: Towards an interpretable deep learning model of cancer

Affiliation: Karolinska Institutet

Bio: Avlant Nilsson is a computational biologist and assistant professor in precision medicine at the department of Cell and Molecular Biology at Karolinska Institutet, Stockholm. He holds a MSc (2009-2014), and a PhD (2014-2019) degree in biological engineering from Chalmers University of Technology, where his thesis focused on the metabolism of growing cells, including liver cancer.
In his postdoctoral work at Massachusetts Institute of Technology (2019-2023), he developed artificial neural network models to simulate signal transduction in immune cells.
His lab at SciLifeLab is developing techniques to simulate cellular processes in cancer, aiming at identifying effective drug combinations, predicting resistance mechanisms, and understanding cell-cell interactions in the tumor microenvironment.

Abstract: Deep learning offers new possibilities for understanding cancer using high throughput data. However, it can be challenging to translate predictive models into causal description of cellular responses. Our lab develops biologically informed neural networks that integrate omics data with prior knowledge of signaling, gene regulation, and metabolism. By constraining the models to only include physical molecular interactions, we are developing interpretable models for computer-aided design of personalized cancer medicine. With this approach we aim to answer questions such as why the same mutation yields different effects across cell types, unexpected signaling outcomes of drugs, and how genetic alterations may drive metabolic strategies.

Title: Climate adaptation in natural forest trees

Affiliation: Umeå Plant Sciences Centre, Swedish Agricultural University

Abstract: Conifers are ecologically dominant and economically important, but are succumbing to drought, disease, early-budding and other challenges globally because the climate has changed so that mature trees are no longer adapted to their environment. If we could predict how individual tree genotypes would respond to different environments, we could — given environmental predictions — plant the right tree in the right space. Standard agronomic approaches are effective but are less suitable for trees with long generation times and huge genetic diversity. We propose a system for quickly estimating adaptive responses for any forest tree. The key is tree increment core samples, which simultaneously provide DNA for genotyping and annual growth measurements, estimated from growth rings. For each genotype, we thus have a life-time’s worth of experienced year-environments. This allows us to partition growth variation into generalizable environmental responses for years with historical weather or biotic information, using quantitative, genomic and ecological approaches to control for correlated responses. We focus on the economically and ecologically important conifer Norway spruce (Picea abies) to 1) develop models and infrastructure to understand the fraction of annual growth that can be attributed to genotype, environment and genotype-by-environment interactions (GxE), 2) map the genetic basis of adaptive response using estimates for GxE as a response in genome-wide association studies (GWAS) and 3) predict genetic responses to novel environments. This approach enables estimation of the genetic basis of adaptive responses in any population, providing the means to evaluate a tree’s performance in any modeled environment. As environments shift under climate change, this provides a powerful tool to select parents for healthy, resilient forests.

Title: Bridging the last mile: microbiome-enhanced forecasting of pregnancy complications

Affiliation: Uppsala University

Bio: Luisa W. Hugerth trained both in biomedicine at Karolinska and in microbial ecology at KTH before combining these two interests by joining the Centre for Translatinal Microbiome Research. Since 2022 she is a DDLS fellow based at UU, where she leads the Human Microbial Ecology Lab, studying the microbiome as a modifiable risk factor in women’s health.

Abstract: Patients know a lot about themselves and their health, and asking the right questions will remain the key main strategy for taking clinical decisions. Still, there is always uncertainty in how an individual will react to an intervention. Leveraging large population-based pregnancy cohorts and combining questionnaires, registries, fecal and vaginal microbiome, we achieve improved accuracy in pregnancy complication predictions, including preterm birth, large-for-gestational age and small-for-gestational age infants.