Biology subjects

Hildebrandt, A.

Publications and source records attributed to Hildebrandt, A..

11 recordsLinked to original sources

bioRxiv2026

Cross-Domain Transfer Learning from Peptides to Lipids Using a Multi-Property Fine-Tuned LLM

Accurate knowledge of liquid chromatography retention time (RT) is essential for confident compound identification in metabolomics and lipidomics. Yet, it is often constrained by the scarcity of experimental data for many molecular classes. Current workflows depend on experimental RT libraries, which are time-consuming to build and limited to previously observed compounds. Here, we present a transfer learning pipeline that leverages large, publicly available peptide datasets to enable accurate lipid RT prediction in data-sparse scenarios. We first show that a ChemBERTa language model, when pre-trained on peptides with a multi-task objective (predicting both RT and fundamental RDKit molecular descriptors), learns a more robust and generalizable chemical representation than a single-task (RT-only) model. This multi-property pre-training yielded superior generalization in lipids, achieving test R2 values of 0.842 against 0.814 (RT-only). Crucially, transferring this peptide-based model to lipid data provided a pronounced advantage in data-sparse scenarios. When fine-tuned on only 5% of available lipid data, the transferred model improved the median test R2 by +0.234 over a model trained from scratch. Significant benefits persisted at intermediate data scales (50-75%), with performance converging only when 100% of the lipid data was used. Notably, the pre-trained model never underperformed the baseline, exhibiting more stable training across all data scales. These results demonstrate that multi-property pre-training guides language models towards chemically meaningful representations that support better RT prediction in different molecular domains. Furthermore, peptide-based pre-training facilitates cross-domain transfer of chemical properties to lipid. Our work provides a practical, scalable strategy to mitigate data scarcity in lipidomics by transferring knowledge from data-rich peptide databases, offering a computational alternative to extensive experimental library generation and enabling more confident identification in small-scale omics studies.

bioinformatics↗

bioRxiv2025

Lost in a Large EEG Multiverse? Comparing Sampling Approaches for Representative Pipeline Selection

The multiplicity of defensible strategies for processing and analysing data has been implicated as a core contributor to the replicability crisis, creating uncertainty about the robustness of a result to variations in data processing choices. This issue is exacerbated where a large number of data processing pipelines are defensible, and where there is great heterogeneity in the pipelines applied in the literature, such as in processing and analysing electroencephalography (EEG) signals. In a multiverse analysis, equally defensible pipelines are computed and the robustness of the result to these variations is reported. However, a large number of defensible pipelines is sometimes infeasible to compute exhaustively, and researchers rely on sampling approaches. In these cases, pipelines are sampled from the full multiverse and the robustness is reported across these samples, assuming that they are representative for the entire multiverse. However, different sampling methods may yield different robustness results, introducing what we term multiverse sampling uncertainty. To illustrate, we computed a 528-pipeline multiverse analysis on EEG-recordings during an emotion classification task aiming to predict extraversion scores from the Late Positive Potential. We applied three sampling methods (random, stratified, and active learning) to sample 26 pipelines (5%), and evaluated the results in terms of the representativeness of the distribution of model fits to that of the full multiverse. Our results highlight variability in the representativeness of the distribution of model fits between samples. The active learning sample most closely represented the median model fit of the full multiverse. The need for representative pipeline sampling to mitigate bias in large multiverse analyses is discussed.

neuroscience↗

bioRxiv2024

The multiverse of data preprocessing and analysis in graph-based fMRI: A systematic literature review of analytical choices fed into a decision support tool for informed analysis

The large number of different analytical choices researchers use may be partly responsible for the replication challenge in neuroimaging studies. For robustness analysis, knowledge of the full space of options is essential. We conducted a systematic literature review to identify the analytical decisions in functional neuroimaging data preprocessing and analysis in the emerging field of cognitive network neuroscience. We found 61 different steps, with 17 of them having debatable options. Scrubbing, global signal regression, and spatial smoothing are among the controversial steps. There is no standardized order in which different steps are applied, and the options within several steps vary widely across studies. By aggregating the pipelines across studies, we propose three taxonomic levels to categorize analytical choices: 1) inclusion or exclusion of specific steps, 2) distinct sequencing of steps, and 3) parameter tuning within steps. To facilitate access to the data, we developed a decision support app with high educational value called METEOR, which allows researchers to explore the space of choices as reference for well-informed robustness (multiverse) analysis. HighlightsO_LIData analysis variability in neuroimaging hinders replicability. C_LIO_LIAnalysis across multiple defensible options examines the robustness of results. C_LIO_LIWe conducted a systematic literature review to identify analytical options. C_LIO_LIWe identified 61 steps and 102 options in performing graph-fMRI analysis. C_LIO_LIInteractive visualization of these steps and options is available as a Shiny app. C_LI

neuroscience↗

bioRxiv2023

Plant diversity stabilizes soil temperature

Extreme weather events are occurring more frequently, and research has shown that plant diversity can help mitigate impacts of climate change by increasing plant productivity and ecosystem stability1,2. Although soil temperature and its stability are key determinants of essential ecosystem processes related to water and nutrient uptake3 as well as soil respiration and microbial activity4, no study has yet investigated whether plant diversity can buffer soil temperature fluctuations. Using 18 years of a continuous dataset with a resolution of 1 minute ([~]795,312,000 individual measurements) from a large-scale grassland biodiversity experiment, we show that plant diversity buffers soil temperature throughout the year. Plant diversity helped to prevent soil heating in hot weather, and cooling in cold weather. Moreover, this effect of plant diversity increased over the 18-year observation period with the aging of experimental communities and was even stronger under extreme conditions, i.e., on hot days or in dry years. Using structural equation modelling, we found that plant diversity stabilized soil temperature by increasing soil organic carbon concentrations and, to a lesser extent, by increasing the plant leaf area index. We suggest that the diversity-induced stabilization of soil temperature may help to mitigate the negative effects of extreme climatic events such as soil carbon release, thus slow global warming.

ecology↗

bioRxiv2023

midiaPASEF maximizes information content in data-independent acquisition proteomics

Data-independent acquisition (DIA) approaches provide comprehensive records of all detectable pre-cursor and fragment ions. Here we introduce midiaPASEF, a novel DIA scan mode using mobility-specific micro-encoding of overlapping quadrupole windows to optimally cover the ion population in the ion mobility-mass to charge plane. Using overlapping ion mobility-encoded quadrupole windows, midiaPASEF maximizes information content in DIA acquisitions which enables the determination of the precursor m/z of each fragment ion with a precision of less than 2 Th. The Snakemake-based MIDIAID pipeline integrates algorithms for multidimensional peak detection and for machine-learning-based classification of precursor-fragment relationships. The MIDIAID pipeline enables fully automated processing and multidimensional deconvolution of midia-PASEF files and exports highly specific DDA-like MSMS spectra which are suitable for de novo sequencing and can be searched directly with established tools including PEAKS, FragPipe and Mascot. midiaPASEF acquisition identifies over 40 unique peptides per second and provides powerful library-free DIA analyses including phosphopeptidome and immunopeptidome samples.

bioinformatics↗

bioRxiv2021

Locality-sensitive hashing enables signal classification in high-throughput mass spectrometry raw data at scale

Mass spectrometry is an important experimental technique in the field of proteomics. However, analysis of certain mass spectrometry data faces a combination of two challenges: First, even a single experiment produces a large amount of multi-dimensional raw data and, second, signals of interest are not single peaks but patterns of peaks that span along the different dimensions. The rapidly growing amount of mass spectrometry data increases the demand for scalable solutions. Existing approaches for signal detection are usually not well suited for processing large amounts of data in parallel or rely on strong assumptions concerning the signals properties. In this study, it is shown that locality-sensitive hashing enables signal classification in mass spectrometry raw data at scale. Through appropriate choice of algorithm parameters it is possible to balance false-positive and false-negative rates. On synthetic data, a superior performance compared to an intensity thresholding approach was achieved. The implementation scaled out up to 88 threads on real data. Locality-sensitive hashing is a desirable approach for signal classification in mass spectrometry raw data. Generated data and code are available at https://github.com/hildebrandtlab/mzBucket. Raw data is available at https://zenodo.org/record/5036526.

bioinformatics↗

bioRxiv2020

Resting state brain signal complexity of young healthy adults reflects genetic risk for developing Alzheimer Disease

The e4 allele of the APOE gene is strongly associated with impaired brain functionality and cognitive decline in humans at older age. It is controversial whether and how the APOE e4 allele is affecting brain activity among young healthy individuals and how such effects may contribute to individual differences in cognitive performance. Signal complexity is a critical aspect of brain activity that has been shown to be associated with brain function. In this study, we analyzed multiscale entropy (MSE) of EEG signals among young healthy adults as an indicator of brain signal complexity and investigated how MSE is predicted by APOE genotype groups. Furthermore, by means of structural equation modeling, we investigated whether MSE predicts fluid intelligence. Results indicate larger MSE in young healthy e4 carriers across all time scales. Moreover, better fluid intelligence (gf) is associated with smaller MSE at low time scales and larger MSE at higher scales. However, MSE does not account for better cognitive performance among APOE e4 carriers by mediating the APOE genotype effect on fluid intelligence. The present results shed further light on the neural mechanisms underlying gene-behavior association relevant for Alzheimers Disease risk.

neuroscience↗

bioRxiv2019

Plant traits are poor predictors of long-term ecosystem functioning

Earth is home to over 350,000 vascular plant species1 that differ in their traits in innumerable ways. Yet, a handful of functional traits can help explaining major differences among species in photosynthetic rate, growth rate, reproductive output and other aspects of plant performance2-6. A key challenge, coined "the Holy Grail" in ecology, is to upscale this understanding in order to predict how natural or anthropogenically driven changes in the identity and diversity of co-occurring plant species drive the functioning of ecosystems7, 8. Here, we analyze the extent to which 42 different ecosystem functions can be predicted by 41 plant traits in 78 experimentally manipulated grassland plots over 10 years. Despite the unprecedented number of traits analyzed, the average percentage of variation in ecosystem functioning that they jointly explained was only moderate (32.6%) within individual years, and even much lower (12.7%) across years. Most other studies linking ecosystem functioning to plant traits analyzed no more than six traits, and when including either only six random or the six most frequently studied traits in our analysis, the average percentage of explained variation in across-year ecosystem functioning dropped to 4.8%. Furthermore, different ecosystem functions were driven by different traits, with on average only 12.2% overlap in significant predictors. Thus, we did not find evidence for the existence of a small set of key traits able to explain variation in multiple ecosystem functions across years. Our results therefore suggest that there are strong limits in the extent to which we can predict the long-term functional consequences of the ongoing, rapid changes in the composition and diversity of plant communities that humanity is currently facing.

ecology↗

bioRxiv2019

The RNA-binding ubiquitin ligase MKRN1 functions in ribosome-associated quality control of poly(A)-translation

Cells have evolved quality control mechanisms to ensure protein homeostasis by detecting and degrading aberrant mRNAs and proteins. A common source of aberrant mRNAs is premature polyadenylation, which can result in non-functional protein products. Translating ribosomes that encounter poly(A) sequences are terminally stalled, followed by ribosome recycling and decay of the truncated nascent polypeptide via the ribosome-associated quality control (RQC). Here, we demonstrate that the conserved RNA-binding E3 ubiquitin ligase Makorin Ring Finger Protein 1 (MKRN1) promotes ribosome stalling at poly(A) sequences during RQC. We show that MKRN1 interacts with the cytoplasmic poly(A)-binding protein (PABP) and is positioned upstream of poly(A) tails in mRNAs. Ubiquitin remnant profiling uncovers PABP and ribosomal protein RPS10, as well as additional translational regulators as main ubiquitylation substrates of MKRN1. We propose that MKRN1 serves as a first line of poly(A) recognition at the mRNA level to prevent production of erroneous proteins, thus maintaining proteome integrity.

molecular biology↗

bioRxiv2018

Makorin1 controls embryonic patterning by alleviating Bruno-mediated repression of oskar translation

Makorins are evolutionary conserved proteins that contain C3H-type zinc finger modules and a RING E3 ubiquitin ligase domain. In Drosophila maternal Makorin 1 (Mkrn1) has been linked to embryonic patterning and germ cell specification. Here, we show that Mkrn1 is required for translational activation of oskar, whose product is critical for axis specification and germ plasm assembly. We demonstrate that Mkrn1 interacts with poly(A) binding protein (pAbp) and binds osk 3 UTR in a region adjacent to A-rich sequences. This binding site also overlaps with Bruno (Bru) responsive elements (BREs), which regulate osk translation. We observe increased association of the translational repressor Bru with osk mRNA upon depletion of Mkrn1, implying that the two proteins compete for osk binding. Consistently, reducing Bru dosage partially rescues viability and Osk protein level in ovaries from Mkrn1 females. We conclude that Mkrn1 controls embryonic patterning and germ cell formation by specifically activating osk translation via displacing Bru from its 3 UTR.

developmental biology↗

bioRxiv2018

GTSF-1 is required for the formation of a functional RNA-dependent RNA Polymerase complex in C. elegans

In every domain of life, Argonaute proteins and their associated small RNAs regulate gene expression. Despite great conservation of Argonaute proteins throughout evolution, many proteins acting in small RNA pathways are not widely conserved. Gametocyte-specific factor 1 (Gtsf1) proteins, characterized by two tandem CHHC zinc fingers and an unstructured, acidic C-terminal tail, are conserved in animals and act in small RNA pathways. In fly and mouse, they are required for fertility and have been shown to interact with Piwi clade Argonautes. We identified T06A10.3 as the Caenorhabditis elegans Gtsf1 homolog and named it gtsf-1. Given its conserved nature and roles in Piwi-mediated gene silencing, we sought out to characterize GTSF-1 in the context of the small RNA pathways of C. elegans. Like its homologs, GTSF-1 is required for normal fertility. Surprisingly, we report that GTSF-1 is not required for Piwi-mediated gene silencing. Instead, gtsf-1 mutants show strong depletion of a class of endogenous small RNAs, known as 26G-RNAs, and fully phenocopy mutants lacking RRF-3, the RNA-dependent RNA Polymerase that synthesizes 26G-RNAs. We show, both in vivo and in vitro, that GTSF-1 specifically and robustly interacts with RRF-3 via its tandem CHHC zinc fingers. Furthermore, we demonstrate that GTSF-1 is required for the assembly of a larger RRF-3 and DCR-1-containing complex, also known as ERIC, thereby allowing for 26G-RNA generation. We propose that GTSF-1 homologs may similarly act to drive the assembly of larger complexes that subsequently act in small RNA production and/or in imposing small RNA-mediated silencing activities.

molecular biology↗