CRC 1371 - Microbiome Signatures: Lecture Series

17.01.2023 - Dr. Heather Armstrong

21.03.2023 - Prof. Dr. Julien Gagneur

Title: Recent algorithms from the TUM chair of Computational Molecular Medicine with relevance for microbial signatures research

21.03.2023 | 4 pm
Prof. Dr. Julien Gagneur

Technical University of Munich
Chair for Computational Molecular Medicine
Boltzmannstr. 3/I
85748 Garching b. München

Venue:
Room 2.98, Gregor-Mendel-Str. 2, 85354 Freising

Host: Dirk Haller

Abstract:

This talk aims to give an overview of methodological advancements from my lab with potential utility for microbial signatures research. I will touch on 3 topics relevant to metagenomics, metaproteomics, and host genetics.

We developed a method that can leverage large genome datasets to identify conserved elements without sequence alignment [1]. We leverage masked language modeling – a pivotal technique in natural language processing – i.e. our model is trained to recover nucleotides artificially masked in the input. We trained the model on over 800 fungal species and demonstrated its utility for sequence motif detection and for deriving state-of-the-art performance predictors for a range of gene regulatory prediction tasks.

Second, I will present Spectralis, an algorithm for de novo peptide sequencing, i.e. reference-free identification of peptides from tandem mass spectrometry data [2]. Significant progress is being made on this long-standing problem thanks to deep learning. Spectralis surpassed 40% sensitivity at 90% precision on spectra for which database search provided a ground truth, nearly doubling state-of-the-art sensitivity.

Third, I will give an overview of methodologies to pinpoint the genetic cause of inherited rare disorders using RNA-seq as a complement to genome sequencing [3-5]. Our tools have helped to improve the diagnostic rate of rare disorders by as much as 15% for certain disorders and are now used by multiple rare-disease consortia.

1. Gankin, Karollus et al. Species-aware DNA language modeling. bioRxiv, 2023
2. Klaproth-Andrade et al. Deep learning-driven fragment ion series classification enables highly precise and sensitive de novo peptide sequencing. bioRxiv, 2023
3. Kremer, Bader et al. Genetic diagnosis of Mendelian disorders via RNA sequencing, Nature communications, 2017
4. Yépez et al. Detection of aberrant events in RNA-seq data. Nature Protocols, 2021
5. Wagner, Çelik, et al. Aberrant splicing prediction across human tissues, Nature genetics, in press

Lecture series 2023

17.01.2023 - Dr. Heather Armstrong

Unfermented β-fructan fibers fuel inflammation in select inflammatory bowel disease patients

21.03.2023 - Prof. Dr. Julien Gagneur

Title: Recent algorithms from the TUM chair of Computational Molecular Medicine with relevance for microbial signatures research

16.05.2023 - Prof. Anneli Peters

Title: coming soon

20.06.2023 - Prof. Dr. Bodo Grimbacher

Title coming soon

25.07.2023 - Prof. Dr. Jörg Vogel

The promises and challenges of programmable RNA antibiotics

26.09.2023 - Prof. Dr. Gianni Panagiotou

Title coming soon

10.10.2023 - PhD David Gomez Varela

Title: coming soon

21.11.2023 - Prof. Vincenzo Barnbara

Title: coming soon