Biology subjects

Frey, B.

Publications and source records attributed to Frey, B..

14 recordsLinked to original sources

bioRxiv2026

Exploring nematode diversity in soils: comparison of extraction methods and sequencing protocols

Soil nematodes play key roles in regulating organic matter decomposition, nutrient cycling and trophic interactions in the soil ecosystem, and are very useful bioindicators of soil condition and soil health. But to characterise them, we need reliable measurements of nematode biodiversity in soils. DNA-based techniques such as metabarcoding have recently explored their biodiversity, with successful results. However, there are many methodological steps that can modify the outcome considerably and require informed decisions. In this study, we tested the effect on nematode alpha and beta diversity metrics of three methodological steps, namely i) type of extraction of genetic material from the matrix, either from soil directly with different starting amount of material or extracting nematodes first via the Baermann or Oostenbrink methods, ii) the choice of taxonomically-relevant DNA segment for amplification, with a nematode specific primer (NemF/18Sr2b) and a universal eukaryotic primer (3NDf/1132rmod), both within 18S rDNA, and iii) the selection of appropriate reference library for taxonomic assignment (comparing three nematode specialised databases and three broad taxonomic databases). Observed differences in alpha diversity were more pronounced when comparing the nematode extraction vs soil extractions. In particular, direct soil extractions co-extracted many non-nematode taxa, especially if amplified with a universal primer, which can significantly reduce the number of nematode reads obtained and therefore the accuracy of nematode community indexes and other nematode-based metrics. The choice of the reference database strongly affected the taxonomic identification, which calls for extreme caution when making this selection. Notably, specialised databases tended to over assign reads to nematodes that were detected as non-nematodes by broad taxonomic databases. However, beta diversity metrics were consistent across methods, highlighting these metrics as robust against methodological choices. Our study clearly demonstrates that the choices of methods for nematode biodiversity studies using molecular tools can have a big impact on the results, and decisions should be made consciously and carefully.

ecology↗

bioRxiv2025

Aridity drives global convergence of desert microbiomes and biogeochemical activities

Deserts cover a third of the worlds surface, supporting unique biomes and ecosystem services. Yet, we lack a comprehensive assessment of what defines and drives the microbial communities that dominate life in these regions. Here, we conducted a standardized field survey in contrasting cold, hot, and polar deserts across the seven continents, and observed geographically distant deserts share similar structure, function, and activities. Desert communities are dominated by genomically streamlined Actinobacteriota and Chloroflexota, and compared with non-desert soils, are significantly enriched with stress tolerance genes, mobile genetic elements, and antiviral strategies, revealing previously unknown ecological and evolutionary dynamics. Metabolically, these communities exhibit reduced capacity for carbohydrate and protein degradation, and instead are enriched for chemosynthetic carbon fixation, continuous energy harvesting using atmospheric trace gases and sunlight, and energy reserve biosynthesis. All sampled soils mediated respiration, trace gas oxidation, and carbon fixation, with detectable activity even in hyper-arid Atacama and Antarctic soils at the margins of life. Driver analyses identified aridity as the primary overriding driver of the microbial communities and biogeochemical activities. Collectively, these findings suggest that aridity selects for metabolically self-sufficient taxa capable of continuously meeting energy and carbon needs independently of vegetation-derived inputs, while enduring physicochemical stressors and potentially elevated viral pressure. These new insights are integral to forecast the future of soils amid increasing desertification. Significance statementDesert soils occupy a vast and expanding portion of Earth, yet what defines and governs their dominant microbial life remains incompletely defined. By assessing the composition, capabilities, and activities of microbial communities across deserts on all seven continents, we identify unifying signatures of life under extreme water limitation. We show microbial communities are highly self-sufficient, capable of acquiring energy and carbon even where plant inputs are minimal. This planetary-scale understanding of the desert microbiome has important ramifications for forecasting potential shifts of microbial communities and the services they provide as desertification intensifies.

microbiology↗

bioRxiv2025

Glycolytic flux sustains human Th1 identity and effector function via STAT1 glycosylation

Fate decisions of T helper (Th) cells are tightly linked to their metabolic states, but precise mechanistic links remain unknown, especially in humans. Using in vitro stimulation in combination with gene editing we studied how metabolic regulation shapes human Th1 cell identity and effector function. Differentiated Th1 cells displayed elevated STAT1 phosphorylation at Tyr701 and Ser727 as well as heightened T-bet and IFN{gamma} expression, which were dampened by CRISPR/Cas9-mediated STAT1 deletion. Metabolic profiling revealed enhanced glycolytic activity in Th1 in comparison to Act.T cells, evidenced by increased extracellular acidification rate, ATP production via glycolysis, glucose uptake, lactate secretion and NADH abundance. SCENITH analysis demonstrated elevated glycolysis-dependent anabolic activity of Th1 cells. Inhibition of glycolysis reduced IFN{gamma} production and STAT1 phosphorylation independent of JAK1/2 activity, STAT1 abundance or SHP-2 activity, implicating glycolysis directly in sustaining STAT1-mediated Th1 functionality. Mechanistically, O-Glycosylation, facilitated by O-Glycosyltransferase, emerged as pivotal in modulating STAT1 activity, as evident through immunoprecipitation and Western blot analysis. Pharmaceutical O-Glycosyltransferase inhibition prevented Th1 differentiation as well as STAT1 O-glycosylation. CRISPR/Cas9 mediated mutation of the O-glycosylation sites Ser499 and Thr510 sites diminished STAT1 Ser727 phosphorylation and IFN{gamma} synthesis. Together, this study highlights glycolysis as key regulator of human Th1 cell identity and effector function, with STAT1 O-Glycosylation selectively maintaining Th1 effector capacity. This mechanism could be explored to safeguard Th1 cells in antiviral immunity and autoimmunity. Graphical abstract O_FIG O_LINKSMALLFIG WIDTH=200 HEIGHT=200 SRC="FIGDIR/small/636644v1_ufig1.gif" ALT="Figure 1"> View larger version (52K): org.highwire.dtl.DTLVardef@1ee58aeorg.highwire.dtl.DTLVardef@1001af6org.highwire.dtl.DTLVardef@1ea2ed7org.highwire.dtl.DTLVardef@1a943fc_HPS_FORMAT_FIGEXP M_FIG C_FIG

immunology↗

bioRxiv2025

Single-cell morphological profiling reveals insights into cell death

Analysis at the single-cell level is a powerful approach to study biological processes and responses to perturbations. However, its application in morphological profiling with phenomics remains underexplored. Here, we use the Cell Painting assay to investigate morphological effects of 53 small molecule compounds, associated with six distinct programmed cell death mechanisms, across six concentrations in MCF7 cells. To compare single-cell and aggregated analysis strategies, we conduct both supervised and unsupervised evaluations aimed at identifying features linked to programmed cell death. We apply an energy distance as a metric to quantify morphological perturbation strength, enabling efficient filtering. Among three tested feature extraction methods, self-supervised DINO embeddings applied to single-cell data captured high-resolution morphological patterns. Focused analyses of apoptosis-inducing compounds revealed biological heterogeneity attributable to specific molecular targets and concentration-dependent effects, which were not apparent in aggregated profiles. In contrast, multi-class classification models for the six programmed cell death mechanisms trained on single-cell features achieved F1 scores of 79.86%, while models trained on aggregated features reached F1 scores of up to 89.97%. Our results highlight the advantages of single-cell data for unsupervised exploration and show that aggregated representations yield more robust and accurate performance in supervised models.

bioinformatics↗

bioRxiv2023

Biodiversity in mountain soils above the treeline

Despite the importance of healthy soils for human livelihood, wellbeing, and safety, current gaps in our knowledge and understanding of biodiversity in soil are numerous, undermining conservation efforts. These gaps are particularly wide in mountain regions where healthy soils are especially important for human safety and yet evidence is accumulating of ongoing degradation, posing significant threats to ecosystem functioning and human settlements. To analyse these gaps in detail, we synthesise current research on the global diversity of microorganisms, cryptogams, and invertebrates in mountain soils above the treeline. This synthesis is based on a semi-quantitative survey of the literature and an expert-based analysis. Our work reveals not only deficiencies in geographic cover but also significant gaps in taxonomic coverage, particularly among soil protists and invertebrates, and a lack of (functional and ecological) description of the uncultivated majority of prokaryotes, fungi, and protists. We subsequently build on this overview to highlight opportunities for research on mountain soils as systems of co-occurring species that interact in complex environmental matrices to fulfil critical functions and make essential contributions to life on land. Closing gaps in biodiversity research in mountain soil is crucial to enhance our understanding and to promote laws and guidelines advancing international soil biodiversity conservation targets in mountains. Addressing sparse and biased data, recognizing the impact of environmental changes on mountain ecosystems, and advocating dedicated policies are essential strategies to safeguard mountain soils and their biodiversity. GLOSSARY O_TBL View this table: org.highwire.dtl.DTLVardef@16682fdorg.highwire.dtl.DTLVardef@121b9feorg.highwire.dtl.DTLVardef@139eadeorg.highwire.dtl.DTLVardef@16abd74org.highwire.dtl.DTLVardef@bc3da9_HPS_FORMAT_FIGEXP M_TBL C_TBL

ecology↗

bioRxiv2023

TCCIA: A Comprehensive Resource for Exploring CircRNA in Cancer Immunotherapy

BackgroundImmunotherapies targeting immune checkpoints have gained increasing attention in cancer treatment, emphasizing the need for predictive biomarkers. Circular RNAs (circRNAs) have emerged as critical regulators of tumor immunity, particularly in the PD-1/PD-L1 pathway, and have shown potential in predicting immunotherapy efficacy. Yet, the detailed roles of circRNAs in cancer immunotherapy are not fully understood. While existing databases focus on either circRNA profiles or immunotherapy cohorts, there is currently no platform that enables the exploration of the intricate interplay between circRNAs and anti-tumor immunotherapy. A comprehensive resource combining circRNA profiles, immunotherapy responses, and clinical outcomes is essential to advance our understanding of circRNA-mediated tumor-immune interactions and to develop effective biomarkers. MethodsTo address these gaps, we constructed the Cancer CircRNA Immunome Atlas (TCCIA), the first database that combines circRNA profiles, immunotherapy response data, and clinical outcomes across multi-cancer types. The construction of TCCIA involved applying standardized preprocessing to the raw sequencing FASTQ files, characterizing circRNA profiles using an ensemble approach based on four established circRNA detection tools, analyzing tumor immunophenotypes, and compiling immunotherapy response data from diverse cohorts treated with immune-checkpoint blockades (ICBs). ResultsTCCIA encompasses over 4,000 clinical samples obtained from 25 cohorts treated with ICBs along with other treatment modalities. The database provides researchers and clinicians with a cloud-based platform that enables interactive exploration of circRNA data in the context of ICB. The platform offers a range of analytical tools, including browse of identified circRNAs, visualization of circRNA abundance and correlation, association analysis between circRNAs and clinical variables, assessment of the tumor immune microenvironment, exploration of tumor molecular signatures, evaluation of treatment response or prognosis, and identification of altered circRNAs in immunotherapy-sensitive and resistant tumors. To illustrate the utility of TCCIA, we showcase two examples, including circTMTC3 and circMGA, by employing analysis of large-scale melanoma and bladder cancer cohorts, which unveil distinct impacts and clinical implications of different circRNA expression in cancer immunotherapy. ConclusionsTCCIA represents a significant advancement over existing resources, providing a comprehensive platform to investigate the role of circRNAs in immuno-oncology. What is already known on this topicPrior knowledge indicated that circRNAs are involved in tumor immunity and have potential as predictive biomarkers for immunotherapy efficacy. However, there lacked a comprehensive database that integrated circRNA profiles and immunotherapy response data, necessitating this study. What this study addsThis study introduces TCCIA, a database that combines circRNA profiles, immunotherapy response data, and clinical outcomes. It provides a diverse collection of clinical samples and an interactive platform, enabling in-depth exploration of circRNAs in the context of checkpoint-blockade immunotherapy. How this study might affect research, practice or policyThe findings of this study offer valuable insights into the roles of circRNAs in tumor-immune interactions and provide a resource for researchers and clinicians in the field of immune-oncology. TCCIA has the potential to guide personalized immunotherapeutic strategies and contribute to future research, clinical practice, and policy decisions in checkpoint-blockade immunotherapy and biomarker development.

bioinformatics↗

bioRxiv2022

Inferring cell motility in complex environments with incomplete tracking data

Cell motility has important influence on cell interactions and functionality for various biological aspects. Deciphering these dynamics often relies on live-cell microscopy measurements, which partly have to deal with limitations that could impair a reliable quantification of their motility. Especially given complex environments and tissue structures, limited observation periods, cells moving in and out of focus and impaired calibration of observation axes often lead to loss of cell tracks and insufficient tracking of motility within several dimensions. However, a reliable quantification of cell motility dynamics is essential when aiming at extrapolating the observed dynamics in order to understand cell population dynamics at larger temporal and spatial scales using appropriate simulation environments. To analyze how incomplete observations affect interpretation and parameterization of cell motility, we combined experimental observations with computational models. Studying individual cell dynamics within 3D collagen environments, we found that the gradual loss of cell tracks leads to an underestimation of several motility parameters with the effect dependent on the collagen density. By extending the automated fitting strategy FitMultiCell to account for cell track loss, we show that we are able to retrieve the actual cell dynamics and, thus, to reliably parameterize cell motility from such incomplete data. Applying our approach to the analysis of CD4+ T cells within 3D collagen environments that were infected with HIV-1, we could show that despite a considerable loss of cell tracks, the data still contained sufficient information to compare individual cell motilities by inferring and simulating their dynamics. Thereby, the analysis allowed us to disentangle the effect of HIV-1 infection and collagen density on individual cell motility. Our extended FitMultiCell-approach presented here provides a solution for the elimination of artifacts from cell track data analysis to robustly infer cell motility dynamics.

systems biology↗

bioRxiv2021

Core and indicative bacterial and fungal taxa define characteristic soil communities of arable land, grassland, and forest

Soil microbial diversity has major influences on ecosystem functions and services. However, due to its complexity and uneven distribution of abundant and rare taxa, quantification of soil microbial diversity remains challenging and thereby impeding its integration into long-term monitoring programs. Using metabarcoding, we analyzed soil bacterial and fungal communities over five years at thirty long-term soil monitoring sites from the three land-use types, arable land, permanent grassland, and forest. Unlike soil microbial biomass and alpha-diversity, microbial community compositions and structures were site- and land-use-specific with CAP reclassification success rates of 100%. The temporally stable site core communities included 38.5% of bacterial and 33.1% of fungal OTUs covering 95.9% and 93.2% of relative abundances. We characterized bacterial and fungal core communities and their land-use associations at the family-level. In general, fungal families revealed stronger land-use type associations as compared to bacteria. This is likely due to a stronger vegetation effect on fungal core taxa, while bacterial core taxa were stronger related to soil properties. The assessment of core communities can be used to form cultivation-independent reference lists of microbial taxa, which may facilitate the development of microbial indicators for soil quality and the use of soil microbiota for long-term soil biomonitoring.

ecology↗

bioRxiv2020

Constitutive androstane receptor directs T cell adaptation to bile acids in the small intestine

Bile acids (BAs) are lipid emulsifying metabolites synthesized in hepatocytes and maintained in vivo through enterohepatic circulation between the liver and small intestine1. As detergents, BAs can cause toxicity and inflammation in enterohepatic tissues2. Nuclear receptors maintain BA homeostasis in hepatocytes and enterocytes3, but it is unclear how mucosal immune cells tolerate high BA concentrations in the small intestine lamina propria (siLP). We previously reported that CD4+ T effector (Teff) cells upregulate expression of the xenobiotic transporter MDR1/ABCB1 in the siLP to prevent BA toxicity and suppress Crohns disease-like small bowel inflammation4. Here, we identify the nuclear xenobiotic receptor, constitutive androstane receptor (CAR/NR1I3), as a regulator of MDR1 expression in T cells, and safeguard against BA toxicity and inflammation in the small intestine. CAR was activated and induced large-scale transcriptional reprograming in Teff cells infiltrating the siLP, but not the colon. CAR induced expression of detoxifying enzymes and transporters in siLP Teff cells, as in hepatocytes, but also the key anti-inflammatory cytokine, Il10. Accordingly, CAR-deficiency in T cells exacerbated, whereas pharmacologic CAR activation suppressed, BA-driven ileitis in T cell-reconstituted Rag-/- mice. These data suggest that CAR acts locally in small intestinal T cells to detoxify BAs and resolve inflammation. Activation of this program offers an unexpected strategy to treat small bowel Crohns disease, and provides evidence of lymphocyte sub-specialization within the small intestine.

immunology↗

bioRxiv2020

Monitoring alpha-synuclein oligomerization and inclusion formation using bimolecular fluorescence complementation assays: What you see is not always what you get.

Bimolecular fluorescence complementation (BiFC) was introduced a decade ago as a method to monitor alpha-synuclein (-syn) oligomerization in intact cells. Since then, several -syn BiFC cellular assays and animal models have been developed based on the assumption that an increase in the fluorescent signal correlates with increased -syn oligomerization or aggregation. Despite the increasing use of these assays and models in mechanistic studies, target validation and drug screening, there have been no reports that 1) validate the extent to which the BiFC fluorescent signal correlates with -syn oligomerization at the biochemical level; 2) provide a structural characterization of the oligomers and aggregates formed by the BiFC fragments; or 3) investigate the extent to which the oligomers of the fluorescent complex resemble oligomers formed on the pathway to -syn fibrillization. To address this knowledge gap, we first analysed the expression level and oligomerization properties of the individual constituents of -syn-Venus, one of the most commonly used BiFC systems, in HEK-293 & SH-SY5Y cells from three different laboratories using multiple approaches, including size exclusion chromatography, semiquantitative Western blot analysis, in-cell crosslinking, immunocytochemistry and sedimentation assays. Next, we investigated the biochemical and aggregation properties of -syn upon co-expression of both BiFC fragments. Our results show that 1) the C-terminal-Venus fused to -syn (-syn-Vc) is present in much lower abundance than its counterpart with N-terminal-Venus fused to -syn (Vn--syn) ; 2) Vn--syn exhibits a high propensity to form oligomers and higher-order aggregates; and 3) the expression of either or both fragments does not result in the formation of -syn fibrils or cellular inclusions. Furthermore, our results suggest that only a small fraction of Vn--syn is involved in the formation of the fluorescent BiFC complex and that some of the fluorescent signal may arise from the association or entrapment of -syn-Vc in Vn--syn aggregates. The fact that the N-terminal fragment exists predominantly in an aggregated state also indicates that one must exercise caution when using this system to investigate -syn oligomerization in cells or in vivo. Altogether, our results suggest that cellular and animal models of oligomerization, aggregation and cell-to-cell transmission that are based on the -syn BiFC systems should be thoroughly characterized at the biochemical level to ensure that they reproduce the process of interest and measure what they are intended to measure. Graphical Abstract O_FIG O_LINKSMALLFIG WIDTH=200 HEIGHT=150 SRC="FIGDIR/small/074161v3_ufig1.gif" ALT="Figure 1"> View larger version (55K): org.highwire.dtl.DTLVardef@fe4896org.highwire.dtl.DTLVardef@4f912forg.highwire.dtl.DTLVardef@1c4e157org.highwire.dtl.DTLVardef@a648e6_HPS_FORMAT_FIGEXP M_FIG C_FIG Bimolecular fluorescence complementation (BiFC) was introduced a decade ago to monitor alpha-synuclein oligomerization in intact cells, based on the assumption that an increase in the fluorescent signal correlates with -synuclein oligomerization and aggregation. Herein, we used several biochemical and cellular assays to characterize commonly used -synuclein Venus BiFC assays. Our results show that one of the BiFC fragments (Vn--synuclein) exhibits higher expression levels and aggregation propensity than its counterpart (-synuclein-Vc), thus complicating the interpretation of the molecular interactions that give rise to the fluorescence signal and raise concerns about their application to investigate -syn oligomerization in cells or in vivo.

biochemistry↗

bioRxiv2020

Ecological patterns of root nodule diversity in cultivated and wild rooibos populations: a community prediction approach

There is interest in understanding the factors behind the biogeography of root-associated bacteria due to the joint effects that plant host, climate, and soil conditions can have on bacterial diversity. For legume crops with remaining wild populations, this is of even more importance, because the effects of cropping on undisturbed root-associated bacterial communities can be addressed. Here, we used a community prediction approach to describe the diversity of the root nodule bacterial communities of rooibos (Aspalathus linearis), an endemic legume crop from South Africa. The goal was to reveal whether patterns of root nodule community composition in paired cultivated and wild rooibos populations could be related to geographical distance, plant traits, and plant population type (i.e. cultivated or uncultivated). We identified a core of dominant and widespread Mesorhizobium ZOTUs that each defined one of 4 different root nodule community classes. Rooibos cultivation impacted root nodule bacterial diversity at regional and local scales, while the geographical origin of the root nodule communities was the strongest predictor of root nodule community structure. Beyond impacts of cultivation on root nodule bacterial diversity, this study suggests a mixture of dispersal limitation and ecological drift regionally, and selection by different plant populations locally, define the biogeography of rooibos root nodule bacterial communities.

ecology↗

bioRxiv2020

Coalescence of rhizobial communities in soil interacts with fertilization and determines the assembly of rhizobia in root nodules

Soil microbial community coalescence, whereby entire microbial communities mix and compete in a new environmental setting, is a widespread phenomenon whose applicability for targeted root microbiome assembly has not been studied. Using a legume shrub adapted to nutrient poor soil, we tested for the first time how the assembly of communities of rhizobial root nodule symbionts is affected by the interaction of coalescence and fertilization. Seedlings of the rooibos [Aspalathus linearis (Burm.f.) Dahlg.], were raised in pairwise mixtures of soil from cultivated and uncultivated land of five farms, as well as the individual mixture components. A fragment of the symbiosis maker gene, nodA, was sequenced to characterize the taxonomic turnover of the rhizobia associated with all root nodules at the age of eight month. Soil mixing promoted taxonomic turnover in the rhizobial communities, while fertilization amplified such turnover by increasing the number of rhizobia that became more abundant after soil mixing. Soil mixing and fertilization had a synergistic effect on the abundance of a particular taxon, which was rare in the component soils but became highly abundant in fertilized plants raised in soil mixtures. These findings provide the first evidence that fertilizer addition can interact with soil microbial community coalescence, probably through increasing the chances for rare strains to prioritize root nodule colonization. The combination of soil mixing and fertilizer addition may be a still unexplored measure to (re)introduce root microbial mutualists in arable land.

ecology↗

bioRxiv2020

Different ecological processes drive the assembly of dominant and rare root-associated bacteria in a promiscuous legume

Understanding how plant-associated microbial communities assemble and the roles they play in plant performance are major goals in microbial ecology. For nitrogen-fixing rhizobia, assemblages are mostly determined by filtering by the host as well as abiotic soil conditions. However, for legumes adapted to highly variable environments and nutrient-poor soils, filtering out rhizobial partners may not be an effective strategy to ensure symbiotic benefits. As a consequence, this can lead to net increases in rhizobial diversity driven by stochastic (neutral) assembly processes. Here, we addressed whether symbiotic promiscuity of rooibos (Aspalathus linearis Burm. Dahlgren), reflects increases in rhizobial diversity that are independent of the environmental conditions, following a neutral assembly. We performed a common garden experiment to assess whether root system size and location- and habitat-specific rhizobial propagule pools of rooibos affected rhizobial community diversity and composition. We found a dominance of drift in driving taxonomic turnover in the root nodule communities, but operating at different scales in the dominant Mesorhizobium symbionts and the rest of bacterial taxa. Specifically, drift triggered differentiation between the core rhizobial symbionts of rooibos due to dispersal limitation on a regional scale, whereas it dominated the assembly of non-dominant rhizobial taxa at the root level. This suggests the existence of a significant neutral component in rhizobial community assembly when selectivity by the host plant is low. We conclude that in this promiscuous legume neutral processes govern bacterial community root nodule community assembly, but that these operate at different scales in dominant and rare rhizobial symbionts of the plant.

ecology↗

bioRxiv2018

cDeepbind: A context sensitive deep learning model of RNA-protein binding

MotivationDetermining RNA binding protein(RBP) binding specificity is crucial for understanding many cellular processes and genetic disorders. RBP binding is known to be affected by both the sequence and structure of RNAs. Deep learning can be used to learn generalizable representations of raw data and has improved state of the art in several fields such as image classification, speech recognition and even genomics. Previous work on RBP binding has either used shallow models that combine sequence and structure or deep models that use only the sequence. Here we combine both abilities by augmenting and refining the original Deepbind architecture to capture structural information and obtain significantly better performance.\n\nResultsWe propose two deep architectures, one a lightweight convolutional network for transcriptome wide inference and another a Long Short-Term Memory(LSTM) network that is suitable for small batches of data. We incorporate computationally predicted secondary structure features as input to our models and show its effectiveness in boosting prediction performance. Our models achieved significantly higher correlations on held out in-vitro test data compared to previous approaches, and generalise well to in-vivo CLIP-SEQ data achieving higher median AUCs than other approaches. We analysed the output from our model for VTS1 and CPO and provided intuition into its working. Our models confirmed known secondary structure preferences for some proteins as well as found new ones where secondary structure might play a role. We also demonstrated the strengths of our model compared to other approaches such as the ability to combine information from long distances along the input.\n\nAvailabilitySoftware and models are available at https://github.com/shreshthgandhi/cDeepbind\n\nContactljlee@psi.toronto.edu, frey@psi.toronto.edu

genomics↗