EMBL Seminars

Abstract
Cells often rely on signalling molecules to make cell-fate decisions during development. We can see these fate-regulating decisions as computational tasks and measure the performance of these tasks with information-theoretic quantifiers. Yet, it is often unclear under what constraints a particular task is performed. Here, I will argue that we can nevertheless learn about the computational performance of a gene regulatory system. First, I will discuss an example in early fly development, where the computational task involves interpreting signals so that cell fates can correctly be distinguished.  I will discuss how clustering of transcription factors can provide a way with which binding sites can achieve optimal information transfer. Second (if there is time), I will focus first on an example from the canonical Wnt pathway, where we explore synthetic signals and infer which signals would need to be present naturally for the computation to be close to optimality.  We find that for appropriately chosen signals, the cellular response can be precise enough to allow reliable differentiation into two distinct states. As the precision in the pathway improves, more distinct states can be reliably distinguished. 

Abstract
Complex heatmaps are efficient to visualize associations between different sources of data sets and reveal potential patterns. Here the ComplexHeatmap R package provides a highly flexible way to arrange multiple heatmaps and supports various annotation graphics.

About the speaker
Zuguang is the creator of ComplexHeatmap, the go-to package for customizable and publication-ready heatmaps in R. He’ll show us some of the powerful features in the package, and share practical tips for visualizing large and complex datasets.

Attachments
https://bioconductor.org/packages/devel/bioc/html/ComplexHeatmap.html

Connection details
Zoom: https://embl-org.zoom.us/j/93966704170?pwd=UE5maWxYbTgvRkVvTU12dkdqUU5CZz09 (Meeting ID: 93966704170, Password: emblr)

Please note that the talk will not be recorded.

Abstract
In situ cryoET data collection requires experienced users and appreciable microscope time to carefully select targets for tilt series acquisition. To facilitate full automation, I developed smart parallel automated cryo-electron tomography (SPACEtomo), a workflow using machine learning approaches for lamella detection and biological feature segmentation followed by automated target setup and parallel acquisition. This degree of automation will be essential for obtaining statistically relevant datasets and high-resolution structures of macromolecules in their native context. Such large datasets subsequently require careful data management and vast processing resources to obtain scientific insights. I will show how DECTRIS CLOUD can be leveraged to gain easy access to virtually unlimited cloud-based computing power for cryoEM and cryoET data processing.

Abstract
My lab has developed lightweight, robust and interpretable deep learning models that can predict diverse biochemical profiles spanning transcription factor binding, chromatin accessibility, nascent transcription, steady state transcription and reporter assays. Our models (1) detect, learn and correct cryptic experimental biases, (2) reveal underlying causal sequence syntax and its pleiotropy across biochemical and cellular contexts, (3) encode biophysical parameters and (4) predict effects of regulatory genetic variants. Our models match or surpass the performance of massive, multi-task supervised models and self-supervised DNA language models across a battery of carefully curated benchmarks, including molecular QTLs from diverse cell contexts and ancestries, reporter assays, CRISPR-based genome editing experiments, and fine mapped GWAS variants. By systematically interpreting a foundational resource of ~5000 regulatory DNA sequence models trained on bulk and single cell datasets from diverse adult and fetal cellular contexts, we can begin unraveling the incredible complexity and context-specificity of regulatory sequence lexicons, syntax and genetic variation encoded in the human genome.

Abstract
TidyMass is a comprehensive computational framework for data processing and analysis of LC-MS data using tidyverse principles. It is developed by the Shen Lab at Nanyang Technological University, Singapore.

About the speaker
Xiaotao develops TidyMass, a modular and tidyverse-friendly toolbox for mass spectrometry-based metabolomics. He’ll introduce the ecosystem behind TidyMass and share his perspective on building reproducible metabolomics workflows in R.

Attachments
https://www.tidymass.org/

Connection details
Zoom: https://embl-org.zoom.us/j/93966704170?pwd=UE5maWxYbTgvRkVvTU12dkdqUU5CZz09 (Meeting ID: 93966704170, Password: emblr)

Please note that the talk will not be recorded.

Abstract
Drastic epigenetic reprogramming occurs during mammalian early embryogenesis. Deciphering the molecular events underlying these processes is crucial for understanding how epigenetic information is transmitted between generations and how life really begins. Probing these questions was previously hindered by the scarce experimental materials that are available in early development. By developing a set of ultra-sensitive chromatin analysis technologies, we investigated chromatin reprogramming during early mouse development for chromatin accessibility, histone modifications, and 3D architecture. These studies unveiled highly dynamic and non-canonical chromatin regulation during maternal-to-zygotic transition and zygotic genome activation (ZGA). However, how ZGA is kickstarted and how the early development program is progressively driven by transcription factors (TFs) remain enigmatic. Recently, we identified key TFs that act at the onset of ZGA, and those that connect ZGA to the first cell fate commitment. In this talk, I will discuss how TFs and epigenetic factors cooperatively establish embryonic program and restore the embryonic epigenomes when the life begins. 

Abstract
Large Language Models (LLMs) like ChatGPT are transforming the way we do science and very prominently how we analyse scientific data. In this talk, I will introduce the fundamentals of LLMs and their emerging multimodal extensions, such as Vision-Language Models (VLMs). Computer scientists seek to enable analysis of biological images using VLMs, turning images into numbers and insights ideally directly. I will discuss current capabilities and limitations of these models for bio-image analysis and highlight the unique potential of LLMs for automating data analysis workflows through code-generation. Emphasizing practical examples, I will demonstrate how LLM-driven code-generation accelerates data exploration, preprocessing, and quantitative analysis in bio-image data science.

Abstract
Memory formation relies on a bidirectional interplay between synaptic plasticity and nucleus-templated transcriptional programs, but how precisely this interplay is orchestrated by epigenetic mechanisms remains to a large extent unknown. In this talk, I will showcase our recent efforts to better understand this aspect from two angles. First, we have found that chromatin plasticity in the mouse brain is a key determinant for memory allocation, the process by which neurons become recruited into the memory trace: When we increased chromatin plasticity by enzymatic overexpression of histone acetyl transferases (HATs), neurons with elevated histone acetylation were preferentially recruited into the encoding ensemble and memory retention was enhanced, while optogenetic silencing of the epigenetically altered neurons prevented memory expression. Second, we have found that after learning, the epigenetic make-up of a single locus in the encoding ensemble is necessary and sufficient to bidirectionally alter memory performance across different phases of memory consolidation. Together, these findings stipulate that before and after memory encoding, epigenetic mechanisms play a pivotal role as molecular memory aids. 

Abstract
[Text for abstract].

About the speaker
[Biographical information about the speaker].

Meet the speaker
To meet with the speaker informally after the talks,sign up here [add link]. We especially encourage predocs and postdocs to take advantage of this opportunity.

Attachments
[Link to a file (for example a pdf of the seminar’s programme) - the file can be uploaded on the intranet]

Connection details
Zoom*: [https://embl-org.zoom.us/j/96374261689?pwd=TnNxRWtQY2lyc2pSa2JpY3NGcDlhZz09] (Meeting ID: [963 7426 1689], Password: [DBU])

Please note that the talk will yes/not be recorded.
*For the FAQ section, as a zoom participant, please use either the chat function (the host will read out your question) or the “raise your hand” function and turn on your microphone.