Biology subjects

Lang, M.

Publications and source records attributed to Lang, M..

At least 37 records · Page 2Linked to original sources

bioRxiv2024

ChatNT: A Multimodal Conversational Agent for DNA, RNA and Protein Tasks

Language models are thriving, powering conversational agents that assist and empower humans to solve a number of tasks. Recently, these models were extended to support additional modalities including vision, audio and video, demonstrating impressive capabilities across multiple domains including healthcare. Still, conversational agents remain limited in biology as they cannot yet fully comprehend biological sequences. On the other hand, high-performance foundation models for biological sequences have been built through self-supervision over sequencing data, but these need to be fine-tuned for each specific application, preventing transfer and generalization between tasks. In addition, these models are not conversational which limits their utility to users with coding capabilities. In this paper, we propose to bridge the gap between biology foundation models and conversational agents by introducing ChatNT, the first multimodal conversational agent with an advanced understanding of biological sequences. ChatNT achieves new state-of-the-art results on the Nucleotide Transformer benchmark while being able to solve all tasks at once, in English, and to generalize to unseen questions. In addition, we have curated a new set of more biologically relevant instructions tasks from DNA, RNA and proteins, spanning multiple species, tissues and biological processes. ChatNT reaches performance on par with state-of-the-art specialized methods on those tasks. We also present a novel perplexity-based technique to help calibrate the confidence of our model predictions. Our framework for genomics instruction-tuning can be easily extended to more tasks and biological data modalities (e.g. structure, imaging), making it a widely applicable tool for biology. ChatNT is the first model of its kind and constitutes an initial step towards building generally capable agents that understand biology from first principles while being accessible to users with no coding background.

genomics↗

bioRxiv2024

SegmentNT: annotating the genome at single-nucleotide resolution with DNA foundation models

Genome annotation models that directly analyze DNA sequences are indispensable for modern biological research, enabling rapid and accurate identification of genes and other functional elements. This capability is paramount as the volume of sequenced genomes rapidly expands, making the need for efficient and accurate annotation methods increasingly critical, particularly in the context of genetic variant prediction and in-silico sequence design. Current annotation tools are typically developed for specific element classes and trained from scratch using supervised learning on datasets that are often limited in size. This approach constrains their performance and ability to generalize to new genomes. Here, we frame the genome annotation problem as instance segmentation and introduce a novel methodology for fine-tuning pre-trained DNA foundation models to segment 14 different genic and regulatory elements at single-nucleotide resolution. We leverage the self-supervised pre-trained model Nucleotide Transformer (NT) to develop a general segmentation model, SegmentNT, capable of processing DNA sequences up to 50kb long. By utilizing pre-trained weights from NT, SegmentNT surpasses the performance of several ablation models and baselines, including convolutional networks with one-hot encoded nucleotide sequences and large models trained from scratch. We demonstrate state-of-the-art performance on gene annotation, splice site and regulatory elements detection throughout the genome. We also leveraged our framework to accommodate two extra DNA foundation models, Enformer and Borzoi, extending the sequence context up to 500kb and enhancing performance on regulatory elements. Finally, we show that a SegmentNT model trained on human genomic elements generalizes to elements of different species, and a multi-species SegmentNT model achieves strong generalization across unseen species. Our approach is readily extensible to additional genomic elements and species. We have made our SegmentNT human and multi-species models, as well as the SegmentEnformer and SegmentBorzoi models, available on our github repository in Jax and HuggingFace space in Pytorch.

genomics↗

bioRxiv2023

Dopamine-iron homeostasis interaction rescues mitochondrial fitness in Parkinson's disease

Imbalances of iron and dopamine metabolism along with mitochondrial dysfunction have been linked to the pathogenesis of Parkinsons disease (PD). We have previously suggested a direct link between iron homeostasis and dopamine metabolism, as dopamine can increase cellular uptake of iron into macrophages thereby promoting oxidative stress responses. In this study, we investigated the interplay between iron, dopamine, and mitochondrial activity in neuroblastoma SH-SY5Y cells and human induced pluripotent stem cell (hiPSC)-derived dopaminergic neurons differentiated from a healthy control and a PD patient with a mutation in the -synuclein (SNCA) gene. In SH-SY5Y cells, dopamine treatment affected the expression of transmembrane iron transporters and cellular iron accumulation. Furthermore, dopamine supplementation led to decreased mitochondrial respiration and reduced mitochondrial fitness, including reduced mtDNA copy number and citrate synthase activity, increased oxidative stress and impaired aconitase activity. In dopaminergic neurons derived from a healthy control individual, dopamine showed comparable effects as observed in SH-SY5Y cells. The hiPSC-derived PD neurons harboring an endogenous SNCA mutation demonstrated altered mitochondrial iron homeostasis, reduced mitochondrial capacity along with increased oxidative stress and alterations of tricarboxylic acid cycle linked metabolic pathways compared with control neurons. Importantly, dopamine treatment of these PD neurons promoted a rescue effect by increasing mitochondrial respiration, activating antioxidant stress response, and normalizing altered metabolite levels linked to mitochondrial function. These observations provide evidence that dopamine affects iron homeostasis, intracellular stress responses and mitochondrial function in healthy cells, while dopamine supplementation can restore this disturbed regulatory network in PD cells.

neuroscience↗

bioRxiv2023

LightMHC: A Light Model for pMHC Structure Prediction with Graph Neural Networks

The peptide-major histocompatibility complex (pMHC) is a crucial protein in cell-mediated immune recognition and response. Accurate structure prediction is potentially beneficial for protein interaction prediction and therefore helps immunotherapy design. However, predicting these structures is challenging due to the sequential and structural variability. In addition, existing pre-trained models such as AlphaFold 2 require expensive computation thus inhibiting high throughput in silico peptide screening. In this study, we propose LightMHC: a lightweight model (2.2M parameters) equipped with attention mechanisms, graph neural networks, and convolutional neural networks. LightMHC predicts full-atom pMHC structures from amino-acid sequences alone, without template structures. The model achieved comparable or superior performance to AlphaFold 2 and ESMFold (93M and 15B parameters respectively), with five-fold acceleration (6.65 seconds/sample for LightMHC versus 36.82 seconds/sample for AlphaFold 2), potentially offering a valuable tool for immune protein structure prediction and immunotherapy design.

bioinformatics↗

bioRxiv2023

FrameDiPT: SE(3) Diffusion Model for Protein Structure Inpainting

Protein structure prediction field has been revolutionised by deep learning with protein folding models such as AlphaFold 2 and ESMFold. These models enable rapid in silico prediction and have been integrated into de novo protein design and protein-protein interaction (PPI) prediction. However, biologically relevant features dependent on conformational distributions cannot be estimated with these models. Diffusion models, a novel class of generative models, have been developed to learn conformational distributions and applied to de novo protein design. Limited work has been done on protein structure inpainting, where a masked section is recovered by simultaneously conditioning on its sequence and the rest of the structure. In this work, we propose FrameDiff inPainTing (FrameDiPT), a generalised model for protein inpainting. This is important for T-cells given the hyper-variability of the complementarity determining region (CDR) loops. We evaluated the model on CDR loop design for T-cell receptors and achieved comparable prediction accuracy to ProteinGenerator and RFdiffusion with limited training data and learnable parameters. Different from deterministic structure prediction models, FrameDiPT captures the conformational distribution at different regions and binding states, highlighting a key advantage of generative models. The model and inference code have been released1.

immunology↗

bioRxiv2023

The Auxin Response Factor ARF27 is required for maize root morphogenesis

Crop root systems are central to nutrition acquisition and water usage. Root hairs and lateral roots contribute to fine-scale patterning of root systems and can offer many advantages for improving root function without drastically impacting overall system architecture. Key genetic regulators underpinning root hair morphogenesis have been well characterized in the model plant Arabidopsis but are less understood in maize. Here, we identify a novel determinant of root hair morphogenesis and auxin responses in maize, AUXIN RESPONSE FACTOR27 (ARF27), using both reverse and quantitative genetic approaches. ARF27 is enriched in maize primary root tissues at both the transcript and protein level. Loss of ARF27 leads to short primary roots and reduced root hair formation, while lateral root density is unaltered. In arf27 roots, auxin-responsive gene expression is dysregulated, which is consistent with the predicted function of this transcription factor. Moreover, a genome wide association study (GWAS) to uncover genetic determinants of auxin-dependent root traits identified ARF27 as a candidate gene. Furthermore, auxin hypersensitive maize genotypes exhibit altered crown root length and surface area in field-grown plants. A gene regulatory network (GRN) was reconstructed and an ARF27 subnetwork was integrated with DAP-seq and GWAS data to identify ARF27 target genes. The ARF27 GRN includes known maize root development genes, such as ROOTLESS CONCERNING CROWN ROOTS (RTCS), ROOTHAIRLESS 3 (RTH3) and RTH6. Altogether this work identifies a novel genetic driver of auxin-mediated root morphogenesis in maize that can inform agricultural strategies for improved crop performance.

plant biology↗

bioRxiv2023

Uridine as a potentiator of aminoglycosides through activation of carbohydrate transporters

Aminoglycosides (AGs) are broad-spectrum antibiotics effective against Gram-negative bacteria. AG uptake depends on membrane potential, but the precise mechanisms are incompletely understood. We report here a new mechanism of active AG uptake in Gram-negative bacteria. In E. coli, overexpression of various carbohydrate transporters increases susceptibility to AGs. Conversely, deletion of a single transporter has little impact. We propose a new uptake model where AGs act as substrates for redundant carbohydrate transporters. This mechanism appears to be shared among Gram-negative ESKAPE pathogens. We screened for molecules that induce transporters expression and identified uridine. When uridine is co-administered with AGs under conditions mimicking urinary tract infections, the efficacy of AG therapies is significantly improved against E. coli, including resistant strains, due to enhanced bacterial uptake. Based on previous knowledge on the use of uridine in humans, we propose that uridine can be a potentiating adjuvant to AG treatment of infectious diseases in the hospital.

microbiology↗

bioRxiv2023

Identification of the active mechanism of aminoglycoside entry in V. cholerae through characterization of sRNA ctrR, regulating carbohydrate utilization and transport.

The possible active entry of aminoglycosides in bacterial cells has been debated since the development of this antibiotic family. Here we report the identification of their active transport mechanism in Vibrio species. We combined genome-wide transcriptional analysis and fitness screens to identify alterations driven by treatment of V. cholerae with sub-minimum inhibitory concentrations (sub-MIC) of the aminoglycoside tobramycin. RNA-seq data showed downregulation of the small non-coding RNA ncRNA586 during such treatment, while Tn-seq revealed that inactivation of this sRNA was associated with improved fitness in the presence of tobramycin. This sRNA is located near sugar transport genes and previous work on a homologous region in Vibrio tasmaniensis suggested that this sRNA stabilizes gene transcripts for carbohydrate transport and utilization, as well as phage receptors. The role for ncRNA586, hereafter named ctrR, in the transport of both carbohydrates and aminoglycosides, was further investigated. Flow cytometry on cells treated with a fluorescent aminoglycoside confirmed the role of ctrR and of carbohydrate transporters in differential aminoglycoside entry. Despite sequence diversity, ctrR showed functional conservation across the Vibrionales. This system in directly modulated by carbon sources, suggesting regulation by carbon catabolite repression, a widely conserved mechanism in Gram-negative bacteria, priming future research on aminoglycoside uptake by sugar transporters in other bacterial species.

microbiology↗

bioRxiv2023

RavA-ViaA links Vibrio cholerae Cpx and Zra2 envelope stress to antibiotic response

RavA-ViaA were reported to play a role in aminoglycoside sensitivity but the mechanisms remain elusive. Here, we performed competition and survival experiments to confirm that deletion of ravA-viaA increases tolerance of the Gram-negative pathogen Vibrio cholerae to low and high aminoglycoside concentrations, during aerobic growth. Using high throughput strategies in this species, we identify Cpx and Zra2 two-component systems as new partners of RavA-ViaA. We show that the aminoglycoside tolerance of{Delta} ravvia requires the presence of these membrane stress sensing two-component systems. We propose that deletion of the RavA-ViaA function facilitates the response aminoglycosides because of a pre-activated state of Cpx and Zra2 membrane stress response systems. We also find an impact of these genes on polymyxin B sensitivity and vancomycin resistance, and we show that simultaneous inactivation of ravvia function together with envelope stress response systems leads to outer membrane permeabilization. Vancomycin is mostly used for Gram-positive because of its low efficiency for crossing the Gram-negative outer membrane. Targeting of the ravA-viaA operon for inactivation could be a future strategy to allow uptake of vancomycin into multidrug resistant Gram-negative bacteria.

microbiology↗

bioRxiv2023

Multiple sequence-alignment-based RNA language model and its application to structural inference

Compared to proteins, DNA and RNA are more difficult languages to interpret because 4-letter-coded DNA/RNA sequences have less information content than 20-letter-coded protein sequences. While BERT (Bidirectional Encoder Representations from Transformers)-like language models have been developed for RNA, they are ineffective at capturing the evolutionary information from homologous sequences because unlike proteins, RNA sequences are less conserved. Here, we have developed an unsupervised Multiple sequence-alignment-based RNA language model (RNA-MSM) by utilizing homologous sequences from an automatic pipeline, RNAcmap. The resulting unsupervised, two-dimensional attention maps and one-dimensional embeddings from RNA-MSM can be directly mapped with high accuracy to 2D base pairing probabilities and 1D solvent accessibilities, respectively. Further fine-tuning led to significantly improved performance on these two downstream tasks over existing state-of-the-art techniques. We anticipate that the pre-trained RNA-MSM model can be fine-tuned on many other tasks related to RNA structure and function.

bioinformatics↗

bioRxiv2022

Peptide-MHC Structure Prediction With Mixed Residue and Atom Graph Neural Network

Neoantigen-targeting vaccines have achieved breakthrough success in cancer immunotherapy by eliciting immune responses against neoantigens, which are proteins uniquely produced by cancer cells. During the immune response, the interactions between peptides and major histocompatibility complexes (MHC) play an important role as peptides must be bound and presented by MHC to be recognised by the immune system. However, only limited experimentally determined peptide-MHC (pMHC) structures are available, and in-silico structure modelling is therefore used for studying their interactions. Current approaches mainly use Monte Carlo sampling and energy minimisation, and are often computationally expensive. On the other hand, the advent of large high-quality proteomic data sets has led to an unprecedented opportunity for deep learning-based methods with pMHC structure prediction becoming feasible with these trained protein folding models. In this work, we present a graph neural network-based model for pMHC structure prediction, which takes an amino acid-level pMHC graph and an atomic-level peptide graph as inputs and predicts the peptide backbone conformation. With a novel weighted reconstruction loss, the trained model achieved a similar accuracy to AlphaFold 2, requiring only 1.7M learnable parameters compared to 93M, representing a more than 98% reduction in the number of required parameters.

bioinformatics↗

bioRxiv2022

Queuosine modification of tRNA-Tyrosine elicits translational reprogramming and enhances growth of Vibrio cholerae with aminoglycosides.

Tgt is the enzyme modifying the guanine (G) in tRNAs with GUN anticodon to queuosine (Q). tgt is required for optimal growth of Vibrio cholerae in the presence of sub-lethal aminoglycoside concentrations. We further explored here the role of the Q34 in the efficiency of codon decoding upon tobramycin exposure. We characterized its impact on the overall bacterial proteome, and elucidated the molecular mechanisms underlying the effects of Q34 modification in antibiotic translational stress response. Using molecular reporters, we showed that Q34 impacts the efficiency of decoding at tyrosine TAT and TAC codons. Proteomics analyses revealed that the anti-SoxR factor RsxA is better translated in the absence of tgt. RsxA displays a codon bias towards tyrosine TAT and overabundance of RsxA leads to decreased expression of genes belonging to SoxR oxidative stress regulon. We also identified conditions that regulate tgt expression. We propose that regulation of Q34 modification in response to environmental cues leads to translational reprogramming of transcripts bearing a biased tyrosine codon usage. In silico analysis further identified candidate genes which could be subject to such translational regulation, among which DNA repair factors. Such transcripts, fitting the definition of modification tunable transcripts, are central in the bacterial response to antibiotics.

microbiology↗

bioRxiv2022

No evidence for asymmetric sperm deposition in a species with asymmetric male genitalia

BackgroundAsymmetric genitalia have repeatedly evolved in animals, yet the underlying causes for their evolution are mostly unknown. The fruitfly Drosophila pachea has asymmetric external genitalia and an asymmetric phallus with a right-sided gonopore. The complex of female and male genitalia is asymmetrically twisted during copulation and males adopt a right-sided copulation posture on top of the female. We wished to investigate if asymmetric male genital morphology and a twisted gentitalia complex may be associated with differential allocation of sperm into female sperm storage organs. MethodsWe examined the internal complex of female and male reproductive organs by micro-computed tomography using Synchrotron X-rays before, during and after copulation. In additon, we monitored sperm aggregation states and timing of sperm transfer during copulation by premature interruption of copulation at different time-points. ResultsThe asymmetric phallus is located at the most caudal end of the female abdomen during copulation. The female reproductive tract, in particular the oviduct, re-arranges during copulation. It is narrow in virgin females and forms a broad vesicle at 20 min after the start of copulation. Sperm transfer into female sperm storage organs (spermathecae) was only in a minority of examined copulation trials (13 / 64). Also, we found that sperm was mainly transferred early, at 2 - 4 min after the start of copulation. We did not detect a particular pattern of sperm allocation in the left or right spermathecae. Sperm adopted a granular or filamentous aggregation state in the female uterus and spermathecae, respectively. DiscussionNo evidence for asymmetric sperm deposition was identified that could be associated with asymmetric genital morphology or twisted complexing of genitalia. Male genital asymmetry may potentially have evolved as a consequence of a complex internal alignment of reproductive organs during copulation in order to optimize low sperm transfer rates.

evolutionary biology↗

bioRxiv2022

Pushing the limits of HiFi assemblies reveals centromere diversity between two Arabidopsis thaliana genomes

Although long-read sequencing can often enable chromosome-level reconstruction of genomes, it is still unclear how one can routinely obtain gapless assemblies. In the model plant Arabidopsis thaliana, other than the reference accession Col-0, all other accessions de novo assembled with long-reads until now have used PacBio continuous long reads (CLR). Although these assemblies sometimes achieved chromosome-arm level contigs, they inevitably broke near the centromeres, excluding megabases of DNA from analysis in pan-genome projects. Since PacBio high-fidelity (HiFi) reads circumvent the high error rate of CLR technologies, albeit at the expense of read length, we compared a CLR assembly of accession Ey15-2 to HiFi assemblies of the same sample performed by five different assemblers starting from subsampled data sets, allowing us to evaluate the impact of coverage and read length. We found that centromeres and rDNA clusters are responsible for 71% of contig breaks in the CLR scaffolds, while relatively short stretches of GA/TC repeats are at the core of >85% of the unfilled gaps in our best HiFi assemblies. Since the HiFi technology consistently enabled us to reconstruct gapless centromeres and 5S rDNA clusters, we demonstrate the value of the approach by comparing these previously inaccessible regions of the genome between two A. thaliana accessions.

genomics↗

bioRxiv2022

A novel mechanism for left-right asymmetry establishment involving tissue remodeling and MyoID

Left-right asymmetries have repeatedly evolved in diverse animals and affect the position, shape or size of specific organs. How novel left-right asymmetries arise remains unknown. Here, we examined Drosophila pachea, where males have evolved unique asymmetric genitalia lobes and a right-sided copulation posture in the past 3-6 million years. We found that male asymmetric lobes grow in pupae during a 360{degrees} clockwise genitalia rotation, a conserved and widespread developmental process in flies. Using two complementary approaches, drug application and a CRISPR-induced MyoID mutant, we altered genitalia rotation and found that asymmetric lobe sizes depend on genitalia rotation completion, while the sidedness of lobe asymmetry is determined by the rotation direction. We then investigated the impact of genital asymmetry on copulation posture. Males with reversed genital asymmetry still mate in the typical right-sided posture, indicating that right-sided behavior is not determined by asymmetric male genitalia. Our study reveals that a novel genitalia asymmetry has evolved through the co-option of a pre-existing tissue remodeling process. Tissue rotation represents a new mechanism, through which a bilateral organ can acquire a left-right size asymmetry. Summary statementHow asymmetric shapes evolve from symmetry is an important question in developmental biology. Using a juvenile hormone derivative, transgenics and CRISPR mutants in the non model species Drosophila pachea, we uncover that the developmental process of male genitalia rotation got co-opted during evolution of D. pachea to determine the extent and direction of a novel genital left-right asymmetry.

evolutionary biology↗

bioRxiv2021

Long-term forest monitoring unravels constant mortality rise in European forests

European forests are an important source for timber production, human welfare, income, protection and biodiversity. During the last two decades, Europe has experienced a number of droughts which were exceptionally within the last 500 years both in terms of duration and intensity and these droughts seem to left remarkable imprints in the mortality dynamics of European forests. However, systematic observations on tree decline with emphasis on single species together with high-resolution drought data has been scarce so far so that deeper insights into mortality dynamics and drought occurrence is still limiting our understanding at continental scale. Here we make use of the ICP Forest crown defoliation dataset, permitting us to retrospectively monitor tree mortality for four major conifers, two major broadleaves as well as a pooled dataset of nearly all minor tree species in Europe. In total, we analysed more than 3 million observations gathered during the last 25 years and employed a high-resolution drought index which is able to assess soil moisture anomaly based on a hydrological water-balance and runoff model every ten days globally. We found significant overall and species-specific increasing trends in mortality rates accompanied by decreasing soil moisture. A generalized linear model identified previous-year soil moisture anomaly as the most important driver of mortality patterns in European forests. Significant interactions appeared between previous-year soil moisture and stand water regime in conifers, strongly suggesting that conifers growing at productive sites are more vulnerable under drought. We conclude that mortality patterns in European forests are currently reaching a concerning upward trend which could be further accelerated by global change-type droughts. Key messageForest mortality has significantly increased over the last 25 years and remained above the long-term mean since 2012.

ecology↗

bioRxiv2021

Developmental timing of Drosophila pachea pupae is robust to temperature changes

Rearing temperature is correlated with the timing and speed of development in a wide range of poikiloterm animals that do not regulate their body temperature. However, exceptions exist, especially in species that live in environments with high temperature extremes or oscillations. Drosophila pachea is endemic to the Sonoran desert in Mexico, in which temperatures and temperature variations are extreme. We wondered if the developmental timing in D. pachea may be sensitive to differing rearing temperatures or if it remains constant. We determined the overall timing of the Drosophila pachea life-cycle at different temperatures. The duration of pupal development was similar at 25{degrees}C, 29{degrees}C and 32{degrees}C, although the relative progress differed at particular stages. Thus, D. pachea may have evolved mechanisms to buffer temperature effects on developmental speed, potentially to ensure proper development and individuals fitness in desert climate conditions.

developmental biology↗

bioRxiv2021

European-wide forest monitoring substantiate the neccessity for a joint conservation strategy to rescue European ash species (Fraxinus spp.)

European ash (Fraxinus excelsior) and narrow-leafed ash (F. angustifolia) are keystone forest tree species in Europe with a broad ecological amplitude and significant economic importance. Besides global warming both species are currently under significant thread by an invasive fungal pathogen that has been progressively spreading throughout the continent for almost three decades. Ash dieback caused by the invasive ascomycete Hymenoscyphus fraxineus is capable of damaging ash trees of all age classes and often leads to the ultimate death of a tree after years of progressively developing crown defoliation. While studies at national and regional level already suggested rapid decline of ash populations as a result of ash dieback, a comprehensive survey at European level with harmonized crown assessment data across countries could shed more light into the population decline from a pan-European perspective and could also pave the way for a new conservation strategy beyond national boarders. Here, we present data from the ICP Forests Level I crown condition monitoring including 27 countries, covering the timespan from 1987-2020. In total, 407 survey plots randomly distributed across these countries were analyzed resulting in >36,000 individual observations. We found a substantial increase in defoliation and mortality over time indicating that crown defoliation has almost doubled during the last three decades. Hotspots of mortality are currently situated in southern Scandinavia and north-eastern Europe, well corresponding to the fact that the disease spread fast from north-east to north-west. Overall survival probability after nearly 30 years of infection has already reached a critical value of 0.51, but with large differences among regions (0.00-0.907). Both a Cox proportional hazard model as well as an Aalen additive regression model strongly suggest that survival of ash is significantly lower in locations with excessive water regime and which experienced more extreme precipitation events during the last two decades. Our results underpin the neccessity for fast governmental acting and joint rescue efforts beyond national boarders since overall mean defoliation will likely reach 50% as early as 2030 as suggested by time series forecasting. We strongly recommend to develop a pan-European conservation strategy before the decline will reach its tipping point resulting into non-reversible loss of diversity in the European forest landscape.

ecology↗