Biology subjects

Smith, K.

Publications and source records attributed to Smith, K..

At least 55 records · Page 3Linked to original sources

bioRxiv2024

An A-T Hoogsteen base pair in a naked DNA hairpin motif: A Protein-Recognized Conformation

In duplex DNA, A-T and G-C form Watson-Crick base pairs, and Hoogsteen pairing only dominates upon protein binding or DNA damage. Using NMR, we show that an A-T Hoogsteen base pair previously observed in crystal structures of transposon DNA hairpins bound to TnpA protein forms in solution even in the absence of TnpA. This Hoogsteen base pair, located adjacent to a dinucleotide apical loop, exists in dynamic equilibrium with a minor Watson-Crick conformation (population [~]11% and lifetime [~]55 {micro}s). Extending the apical loop to three residues inverted the equilibrium, making Watson-Crick the dominant state and the Hoogsteen conformation recognized by TnpA a minor state (population [~]14% and lifetime [~]28 {micro}s). The propensity for Hoogsteen pairing depended on apical loop residues, which form contacts directly or indirectly stabilizing the Hoogsteen conformation. A structure survey did not reveal Hoogsteen pairing near RNA apical loops making them unique to DNA. Our results demonstrate that Hoogsteen can be the dominant state even in naked unmodified duplex DNA and identify 5-CTT(T/C)AG-3 as a DNA-specific apical loop motif stabilized by Hoogsteen pairing. Hoogsteen base pairs may be prevalent in DNA hairpins forming during replication and transcription, with broad implications for the genomic landscape.

biochemistry↗

bioRxiv2024

Highly recurrent multi-nucleotide mutations in SARS-CoV-2

Multi-nucleotide mutations (MNMs) simultaneously replace multiple nu-cleotides. They are a significant contributor to evolution and disease, as well as to misdiagnosis, misannotation and other biases in genome data analysis. MNMs are generally thought to be rare and random events. However, by processing millions of publicly shared genomes, we show that certain MNMs are highly recurrent in SARS-CoV-2: they repeatedly and consistently modify the same multiple nucleotides at the same genome position in the same way. The most frequent of these MNMs have independently occurred hundreds of times across all SARS-CoV-2 lineages. The vast majority of these recurrent MNMs are linked to transcription regulatory sequences. We propose a mechanism that explains them through template switching as part of the natural transcription process of the virus. This previously unknown mutational pattern increases our understanding of the evolution of SARS-CoV-2 and potentially many other nidoviruses. It also has important consequences for computational evolutionary biology: we show that for example recurrent MNMs cause approximately 14% of false positives during inference of recombination in SARS-CoV-2.

evolutionary biology↗

bioRxiv2024

Activation of Wnt/β-catenin signalling by mutually exclusive FBXW11 and CTNNB1 hotspot mutations drives salivary gland basal cell adenoma

Wnt signalling must be just right to promote tumour growth. Basal cell adenoma (BCA) and basal cell adenocarcinoma (BCAC) of the salivary gland are rare tumours that can be difficult to distinguish from each other and other salivary gland tumour subtypes. Due to their rarity, the genetic profiles of BCA and BCAC have not been extensively explored. Using whole-exome and transcriptome sequencing of BCA and BCAC cohorts, we identify a novel recurrent FBXW11 missense mutation (p.F517S) in BCA, that was mutually exclusive with the previously reported CTNNB1 p.I35T gain-of-function (GoF) mutation. These driver events collectively accounted for 94% of BCAs. In vitro, mutant FBXW11 had a dominant negative affect, characterised by defective binding to {beta}-catenin and the accumulation of {beta}-catenin in cells. This was consistent with the nuclear expression of {beta}-catenin observed in BCA cases harbouring the FBXW11 p.F517S mutation and activation of the Wnt/{beta}-catenin pathway. The genomic profiles of BCAC were distinct from BCA, with hotspot DICER1 and HRAS mutations and putative driver mutations affecting PI3K/AKT and NF-{kappa}B signalling pathway genes. A single BCAC, which may represent a malignant transformation of BCA, harboured the recurrent FBXW11 mutation. These findings have important implications for the diagnosis and treatment of BCA and BCAC, which, despite histopathologic overlap, may be unrelated entities.

cancer biology↗

bioRxiv2024

Recon8D: A metabolic regulome network from oct-omics and machine learning

To explore multiomic regulation of the metabolome, we used machine learning to predict metabolomic variation across [~]1000 different cancer cell lines with matched omics data from 8 biomolecular classes: genomics (copy-number and mutations), epigenomics (histone post-translational modifications (PTMs) and DNA-methylation), transcriptomics and RNA splice variants, non-coding transcriptomics (miRNA and lncRNA), proteomics, and phosphoproteomics. Overall, the metabolome is tightly associated with the transcriptome, with coding and non-coding RNAs emerging as top predictors. Peripheral metabolites are predictable via levels of corresponding enzymes, while those in central metabolism require combinatorial predictors in signaling and redox pathways, and may not reflect corresponding pathway expression. We reconstruct multiomic interaction subnetworks for highly predictable metabolites, and YAP1 signaling emerged as a top global predictor across 4 omic layers. We prioritize predictive multiomic features for single-cell and spatial metabolomics assays. Top predictors were enriched for synthetic-lethal interactions and synergistic combination therapies that target compensatory metabolic modulators.

systems biology↗

bioRxiv2024

Temperature drives seagrass recovery across the Western North Atlantic

Climate-driven shifts in herbivores, temperature and nutrient runoff threaten coastal ecosystem resilience. However, our understanding of ecological resilience, particularly for foundation species, remains limited due to a rarity of field experiments that are conducted across appropriate spatial and temporal scales and that investigate multiple stressors. This study aimed to evaluate the resilience of a widespread tropical marine plant (turtlegrass) to disturbances across its geographic range and how this is impacted by environmental gradients in (a)biotic factors. We assessed the resilience (i.e. recovery) of turtlegrass to a simulated disturbance (complete above- and belowground biomass removal) over a year. Contrary to temperate studies, higher temperature generally enhanced seagrass recovery. While nutrients and light availability had minimal impact, combined high levels of nutrients and herbivore grazing (meso and megaherbivore) reduced aboveground recovery. Our results suggest that the resilience of some tropical species, especially in cooler subtropical waters, may initially increase with warming.

ecology↗

bioRxiv2024

Insights into non-crossover recombination from long-read sperm sequencing

Meiotic recombination is a fundamental process that generates genetic diversity by creating new combinations of existing alleles. Although human crossovers have been studied at the pedigree, population and single-cell level, the more frequent non-crossover events that lead to gene conversion are harder to study, particularly at the individual level. Here we show that single high-fidelity long sequencing reads from sperm can capture both crossovers and non-crossovers, allowing effectively arbitrary sample sizes for analysis from one male. Using fifteen sperm samples from thirteen donors we demonstrate variation between and within donors for the rates of different types of recombination. Intriguingly, we observe a tendency for non-crossover gene conversions to occur upstream of nearby PRDM9 binding sites, whereas crossover locations have a slight downstream bias. We further provide evidence for two distinct non-crossover processes. One gives rise to the vast majority of non-crossovers with mean conversion tract length under 50bp, which we suggest is an outcome of standard PRDM9-induced meiotic recombination. In contrast [~]2% of non-crossovers have much longer mean tract length, and potentially originate from the same process as complex events with more than two haplotype switches, which is not associated with PRDM9 binding sites and is also seen in somatic cells.

genetics↗

bioRxiv2024

Compressive Pangenomics Using Mutation-Annotated Networks

Pangenomics is an emerging field that uses a collection of genomes of a species instead of a single reference genome to overcome reference bias and study the within-species genetic diversity. Future pangenomics applications will require analyzing large and ever-growing collections of genomes. Therefore, the choice of data representation is a key determinant of the scope, as well as the computational and memory performance of pangenomic analyses. Current pangenome formats, while capable of storing genetic variations across multiple genomes, fail to capture the shared evolutionary and mutational histories among them, thereby limiting their applications. They are also inefficient for storage, and therefore face significant scaling challenges. In this manuscript, we propose PanMAN, a novel data structure that is information-wise richer than all existing pangenome formats - in addition to representing the alignment and genetic variation in a collection of genomes, PanMAN represents the shared mutational and evolutionary histories inferred between those genomes. By using "evolutionary compression", PanMAN achieves 5.2 to 680-fold compression over other variation-preserving pangenomic formats. PanMANs relative performance generally improves with larger datasets and it is compatible with any method for inferring phylogenies and ancestral nucleotide states. Using SARS-CoV-2 as a case study, we show that PanMAN offers a detailed and accurate portrayal of the pathogens evolutionary and mutational history, facilitating the discovery of new biological insights. We also present panmanUtils, a software toolkit that supports common pangenomic analyses and makes PanMANs interoperable with existing tools and formats. PanMANs are poised to enhance the scale, speed, resolution, and overall scope of pangenomic analyses and data sharing.

bioinformatics↗

bioRxiv2024

Identification of distinct genotypes in circulating RSV A strains based on variants on the virus replication-associated genes

Respiratory syncytial virus is a common cause of respiratory infection that often leads to hospitalization of infected younger children and older adults. RSV is classified into two strains, A and B, each with several subgroups or genotypes. One issue with the definition of these subgroups is the lack of a unified method of identification or genotyping. We propose that genotyping strategies based on the genes coding for replication-associated proteins could provide critical information on the replication capacity of the distinct subgroup, while clearly distinguishing genotypes. Here, we analyzed the virus replication-associated genes N, P, M2, and L from de novo assembled RSV A sequences obtained from 31 newly sequenced samples from hospitalized patients in Philadelphia and 78 additional publicly available sequences from different geographic locations within the US. In-depth analysis and annotation of the protein variants in L and the other replication-associated proteins N, P, M2-1, and M2-2 identified the polymerase protein L as a robust target for genotyping RSV subgroups. Importantly, our analysis revealed non-synonymous variations in L that were consistently accompanied by conserved changes in its co-factor P or the M2-2 protein, suggesting associations and interactions between specific domains of these proteins. These results highlight L as an alternative to other RSV genotyping targets and demonstrate the value of in-depth analyses and annotations of RSV sequences as it can serve as a foundation for subsequent in vitro and clinical studies on the efficiency of the polymerase and fitness of different virus isolates. IMPORTANCEGiven the historical heterogeneity of Respiratory Syncytial Virus (RSV) and the disease it causes, there is a need to understand the circulating RSV strains each season. This requires an informative and consensus method of genotype definition. Here, we carried out a variant analysis that shows a pattern of specific variations among the replication-associated genes of RSV A across different seasons. Interestingly, these variation patterns point to previously defined interactions of domains within these genes, suggesting co-variation in the replication-associated genes. Our results also suggest a genotyping strategy that can prove to be particularly important in understanding the genotype-phenotype correlation in the era of RSV vaccination, where selective pressure on the virus to evolve is anticipated. More importantly, the categorization of RSV strains based on these patterns may prove to be of prognostic value.

microbiology↗

bioRxiv2024

Timing of standard chow exposure determines the variability of mouse phenotypic outcomes and gut microbiota profile

Standard chow diet contributes to reproducibility in animal model experiments since chows differ in nutrient composition, which can independently influence phenotypes. However, there is little evidence of the role of timing in the extent of variability caused by chow exposure. Here, we measured the impact of diet (5V5M, 5V0G, 2920X, and 5058) and timing of exposure (adult exposure (AE), lifetime exposure (LE), and developmental exposure (DE)) on growth & development, metabolic health indicators, and gut bacterial microbiota profiles across genetically identical C57BL6/J mice. Diet drove differences in macro-and micronutrient intake for all exposure models. AE had no effect on measured outcomes. However, LE mice exhibited significant sex-dependent diet effects on growth, body weight, and body composition. LE effects were mostly absent in the DE model, where mice were exposed to chow differences from conception to weaning. Both AE and LE models exhibited similar diet-driven beta diversity profiles for the gut bacterial microbiota, with 5058 diet driving the most distinct profile. Diet-induced beta diversity profiles were sex-dependent for LE mice. Compared to AE, LE drove 9X more diet-driven differentially abundant genera, majority of which were the result of inverse effects of 2920X and 5058. Our findings demonstrate that lifetime exposure to different chow diets has the greatest impact on reproducibility of experimental measures that are common components of preclinical mouse model studies. Importantly, weaning DE mice onto a uniform diet is likely an effective way to reduce unwanted phenotypic variability among experimental models.

developmental biology↗

bioRxiv2023

Topographical and cell type-specific connectivity of rostral and caudal forelimb corticospinal neuron populations

Corticospinal neurons (CSNs) synapse directly on spinal neurons, a diverse group of neurons with unique structural and functional properties necessary for body movements. CSNs modulating forelimb behavior fractionate into caudal forelimb area (CFA) and rostral forelimb area (RFA) motor cortical populations. Despite their prominence, no studies have mapped the diversity of spinal cell types targeted by CSNs, let alone compare CFA and RFA populations. Here we use anatomical and RNA-sequencing methods to show that CSNs synapse onto a remarkably selective group of spinal cell types, favoring inhibitory populations that regulate motoneuron activity and gate sensory feedback. CFA and RFA CSNs target similar spinal cell types, with notable exceptions that suggest these populations differ in how they influence behavior. Finally, axon collaterals of CFA and RFA CSNs target similar brain regions yet receive surprisingly divergent inputs. These results detail the rules of CSN connectivity throughout the brain and spinal cord for two regions critical for forelimb behavior.

neuroscience↗

bioRxiv2023

Convergent evolution of complex adaptive traits enabled human life at high altitudes

Convergent adaptations represent paradigmatic examples of the capacity of natural selection to influence organisms biology. However, the possibility to investigate genetic determinants underpinning convergent complex adaptive traits has been offered only recently by methods for inferring polygenic adaptations from genomic data. Relying on this approach, we demonstrate how high-altitude Andean human groups experienced pervasive selective events at angiogenic pathways, which resemble those previously attested for Himalayan populations despite partial convergence at the single-gene level was observed. This provides unprecedented evidence for the drivers of convergent evolution of enhanced blood perfusion in populations exposed to hypobaric hypoxia for thousands of years.

evolutionary biology↗

bioRxiv2023

Identification of tumor-intrinsic drivers of immune exclusion in acral melanoma

BackgroundAcral melanoma (AM) has distinct characteristics as compared to cutaneous melanoma and exhibits poor response to immune checkpoint inhibitors (ICI). Tumor-intrinsic mechanisms of immune exclusion have been identified in many cancers but less studied in AM. MethodsWe characterized clinically annotated tumors from patients diagnosed with AM at our institution in correlation with ICI response using whole transcriptome RNAseq, whole exome sequencing, CD8 immunohistochemistry, and multispectral immunofluorescence imaging. A defined interferon-{gamma}-associated T cell-inflamed gene signature was used to categorize tumors into non-T cell-inflamed and T cell-inflamed phenotypes. In combination with AM tumors from two published studies, we systematically assessed the immune landscape of AM and detected differential gene expression and pathway activation in a non-T cell-inflamed tumor microenvironment (TME). Two single-cell(sc) RNAseq AM cohorts and 11 bulk RNAseq cohorts of various tumor types were used for independent validation on pathways associated with lack of ICI response. In total, 892 specimens were included in this study. Results72.5% of AM tumors showed low expression of the T cell-inflamed gene signature, with 23.9% of total tumors categorized as the non-T cell-inflamed phenotype. Patients of low CD3+CD8+PD1+ intratumoral T cell density showed poor prognosis. We identified 11 oncogenic pathways significantly upregulated in non-T cell-inflamed relative to T cell-inflamed TME shared across all three acral cohorts (MYC, HGF, MITF, VEGF, EGFR, SP1, ERBB2, TFEB, SREBF1, SOX2, and CCND1). scRNAseq analysis revealed that tumor cell-expressing pathway scores were significantly higher in low vs high T cell-infiltrated AM tumors. We further demonstrated that the 11 pathways were enriched in ICI non-responders compared to responders across cancers, including acral melanoma, cutaneous melanoma, triple-negative breast cancer, and non-small cell lung cancer. Pathway activation was associated with low expression of interferon stimulated genes, suggesting suppression of antigen presentation. Across the 11 pathways, fatty acid synthase and CXCL8 were unifying downstream target molecules suggesting potential nodes for therapeutic intervention. ConclusionsA unique set of pathways is associated with immune exclusion and ICI resistance in AM. These data may inform immunotherapy combinations for immediate clinical translation.

cell biology↗

bioRxiv2023

AlphaFold2 modeling and molecular dynamics simulations of an intrinsically disordered protein

We use AlphaFold2 (AF2) to model the monomer and dimer structures of an intrinsically disordered protein (IDP), Nvjp-1, assisted by molecular dynamics (MD) simulations. We observe relatively rigid dimeric structures of Nvjp-1 when compared with the monomer structures. We suggest that protein conformations from multiple AF2 models and those from MD trajectories exhibit a coherent trend: the conformations of an IDP are deviated from each other and the conformations of a well-folded protein are consistent with each other. We use a residue-residue interaction network (RIN) derived from the contact map which show that the residue-residue interactions in Nvjp-1 are mainly transient; however, those in a well-folded protein are mainly persistent. Despite the variation in 3D shapes, we show that the AF2 models of both disordered and ordered proteins exhibit highly consistent profiles of the pLDDT (predicted local distance difference test) scores. These results indicate a potential protocol to justify the IDPs based on multiple AF2 models and MD simulations. O_FIG O_LINKSMALLFIG WIDTH=200 HEIGHT=106 SRC="FIGDIR/small/551023v1_ufig1.gif" ALT="Figure 1"> View larger version (22K): org.highwire.dtl.DTLVardef@65c7f1org.highwire.dtl.DTLVardef@29cb87org.highwire.dtl.DTLVardef@1f126d2org.highwire.dtl.DTLVardef@12bb76d_HPS_FORMAT_FIGEXP M_FIG The TOC figure: C_FIG

biophysics↗

bioRxiv2023

Cell-type specific molecular signatures of aging revealed in a brain-wide transcriptomic cell-type atlas

Biological aging can be defined as a gradual loss of homeostasis across various aspects of molecular and cellular function. Aging is a complex and dynamic process which influences distinct cell types in a myriad of ways. The cellular architecture of the mammalian brain is heterogeneous and diverse, making it challenging to identify precise areas and cell types of the brain that are more susceptible to aging than others. Here, we present a high-resolution single-cell RNA sequencing dataset containing [~]1.2 million high-quality single-cell transcriptomic profiles of brain cells from young adult and aged mice across both sexes, including areas spanning the forebrain, midbrain, and hindbrain. We find age-associated gene expression signatures across nearly all 130+ neuronal and non-neuronal cell subclasses we identified. We detect the greatest gene expression changes in non-neuronal cell types, suggesting that different cell types in the brain vary in their susceptibility to aging. We identify specific, age-enriched clusters within specific glial, vascular, and immune cell types from both cortical and subcortical regions of the brain, and specific gene expression changes associated with cell senescence, inflammation, decrease in new myelination, and decreased vasculature integrity. We also identify genes with expression changes across multiple cell subclasses, pointing to certain mechanisms of aging that may occur across wide regions or broad cell types of the brain. Finally, we discover the greatest gene expression changes in cell types localized to the third ventricle of the hypothalamus, including tanycytes, ependymal cells, and Tbx3+ neurons found in the arcuate nucleus that are part of the neuronal circuits regulating food intake and energy homeostasis. These findings suggest that the area surrounding the third ventricle in the hypothalamus may be a hub for aging in the mouse brain. Overall, we reveal a dynamic landscape of cell-type-specific transcriptomic changes in the brain associated with normal aging that will serve as a foundation for the investigation of functional changes in the aging process and the interaction of aging and diseases.

neuroscience↗

bioRxiv2023

Adaptive therapy achieves long-term control of chemotherapy resistance in high grade ovarian cancer

Drug resistance results in poor outcomes for most patients with metastatic cancer. Adaptive Therapy (AT) proposes to address this by exploiting presumed fitness costs incurred by drug-resistant cells when drug is absent, and prescribing dose reductions to allow fitter, sensitive cells to re-grow and re- sensitise the tumour. However, empirical evidence for treatment-induced fitness change is lacking. We show that fitness costs in chemotherapy-resistant ovarian cancer cause selective decline and apoptosis of resistant populations in low-resource conditions. Moreover, carboplatin AT caused fluctuations in sensitive/resistant tumour population size in vitro and significantly extended survival of tumour-bearing mice. In sequential blood-derived cell-free DNA and tumour samples obtained longitudinally from ovarian cancer patients during treatment, we inferred resistant cancer cell population size through therapy and observed it correlated strongly with disease burden. These data have enabled us to launch a multicentre, phase 2 randomised controlled trial (ACTOv) to evaluate AT in ovarian cancer.

cancer biology↗

bioRxiv2023

Spatially-resolved profiling of the T-cell receptor repertoire in archival clinical specimens using FUME-TCRseq

Genomic analysis of the T-cell receptor (TCR) reveals the strength, breadth and clonal dynamics of the adaptive immune response to pathogens or cancer. The diversity of the TCR repertoire, however, means that sequencing is technically challenging, particularly for samples with low quality, degraded nucleic acids. Here, we have developed and validated FUME-TCRseq, a robust and sensitive RNA-based TCR sequencing methodology that is suitable for formalin-fixed paraffin-embedded samples and low amounts of input material. FUME-TCRseq incorporates unique molecular identifiers into each molecule of cDNA, allowing correction for sequencing errors and PCR bias. We used RNA extracted from colorectal and head and neck cancers to benchmark the accuracy and sensitivity of FUME-TCRseq against existing methods, and found excellent concordance between the datasets. Furthermore, FUME-TCRseq detected more clonotypes than a commercial RNA-based alternative, with shorter library preparation time and significantly lower cost. The high sensitivity and the ability to sequence RNA of poor quality and limited amount enables quantitative analysis of small numbers of cells from archival tissue sections, which is not possible with other methods. To demonstrate this we performed spatially-resolved FUME-TCRseq of colorectal cancers using macrodissected archival samples, revealing the shifting T-cell landscapes at the transition to an invasive phenotype, and between tumour subclones containing distinct driver alterations. In summary, FUME-TCRseq represents an accurate, sensitive and low-cost tool for the characterisation of T-cell repertoires, particularly in samples with low quality RNA that have not been accessible using existing methodology.

immunology↗

bioRxiv2023

A machine vision guided robot for fully automated embryonic microinjection

Microinjection is a widely used technique for transgenesis, mutagenesis, cell labeling, cryopreservation, and in vitro fertilization in multiple single and multicellular organisms. Microinjection requires specialized skills acquired for each target organism and involves rate limiting and labor-intensive preparatory steps. Here we constructed a machine vision (MV) guided generalized robot that fully automates the process of microinjection in fruit fly (Drosophila melanogaster) and zebrafish (Danio rerio) embryos. The robot uses machine learning (ML) models trained to detect individual embryos in images of agar plates, and models trained to identify specific anatomical locations within each embryo in 3D space using dual view microscopes. The robot uses this information to serially perform microinjection in each detected embryo without any human intervention. We constructed and used three such robots to automatically microinject tens of thousands of Drosophila and zebrafish embryos. We systematically optimized robotic microinjection for each species and validated the use of the robot by performing routine transgenesis with proficiency comparable to highly skilled human practitioners while achieving up to 4x increases in microinjection throughput in Drosophila. The automated microinjection robot was utilized to microinject pools of over 20,000 uniquely barcoded plasmids into 1,713 embryos in two days to rapidly generate more than 400 unique transgenic Drosophila lines. This high throughput microinjection experiment enabled a novel measurement of the number of independent germline integration events per successfully injected embryo. Finally, we showed that robotic microinjection of cryoprotective agents in zebrafish embryos significantly improves vitrification rates and survival of cryopreserved embryos post-thaw as compared to manual microinjection, opening the tantalizing possibility of large-scale cryobanking of aquatic species at an industrial scale. We anticipate that this versatile automated microinjection system can be applied to carry out microinjection for genome-wide manipulation and cryopreservation at scale in a wide range of other organisms.

genetics↗

bioRxiv2023

Cell Painting-based bioactivity prediction boosts high-throughput screening hit-rates and compound diversity

Efficiently identifying bioactive compounds towards a target of interest remains a time- and resource-intensive task in early drug discovery. The ability to accurately predict bioactivity using morphological profiles has the potential to rationalize the process, enabling smaller screens of focused compound sets. Towards this goal, we explored the application of deep learning with Cell Painting, a high-content image-based assay, for compound bioactivity prediction in early drug screening. Combining Cell Painting data and unrefined single-concentration activity readouts from high-throughput screening (HTS) assays, we investigated to what degree morphological profiles could predict compound activity across a set of 140 unique assays. We evaluated the performance of our models across different target classes, assay technologies, and disease areas. The predictive performance of the models was high, with a tendency for better predictions on cell-based assays and kinase targets. The average ROC-AUC was 0.744 with 62% of assays reaching [≥]0.7, 30% reaching [≥]0.8 and 7% reaching [≥]0.9 average ROC-AUC, outperforming commonly used structure-based predictions in terms of predictive performance and compound structure diversity. In many cases, bioactivity prediction from Cell Painting data could be matched using brightfield images rather than multichannel fluorescence images. Experimental validation of our predictions in follow-up assays confirmed enrichment of active compounds. Our results suggest that models trained on Cell Painting data can predict compound activity in a range of high-throughput screening assays robustly, even with relatively noisy HTS assay data. With our approach, enriched screening sets with higher hit rates and higher hit diversity can be selected, which could reduce the size of HTS campaigns and enable primary screening with more complex assays.

bioinformatics↗