Biology subjects

Taylor, C.

Publications and source records attributed to Taylor, C..

At least 19 recordsLinked to original sources

bioRxiv2026

Identification of Cysteine-Reactive Early Mitotic Inhibitors

Cancer remains a pervasive health problem for millions of Americans, and defining effective anticancer therapeutics remains a major obstacle to the long-term survival of patients. Recent advances in novel chemistries and chemical libraries have enabled researchers to discover and develop covalent inhibitors to treat various diseases, including cancer. This has spurred the generation of cysteine-targeting covalent compound libraries that have not been fully explored within the context of the cell cycle and cell division, which are attractive targets for inhibiting cancer proliferation. We report a novel, cell-based high-content chemical screening strategy for discovering early mitotic spindle assembly inhibitors. We coupled this strategy with a cysteine-reactive covalent compound library containing 3,520 molecules to identify novel early mitotic inhibitors. These inhibitors included a microtubule depolymerizer, a mitotic spindle positioning disruptor, and compounds affecting mitotic spindle structure and chromosome alignment. Two of these compounds triggered mitotic arrest and reduced cell viability at nanomolar concentrations. These inhibitors serve as chemical probes to investigate mechanisms of early cell division and to develop novel cancer therapeutics.

biochemistry↗

bioRxiv2026

The ModelSEED Biochemistry Database, 2026 update: grading multi-source thermodynamics

The ModelSEED Biochemistry Database (https://modelseed.org) supplies foundational mass- and charge-balanced reaction networks for metabolic reconstructions. Here we present an update on the biochemistry where the database has been expanded to [~] 46,000 compounds, and [~] 56,000 reactions, featuring [~] 37,000 metabolic structures. We have expanded our approach for handling thermodynamic data, enabling multiple sources of data to be derived, integrated, and presented to the wider research community. We now publish predictions of pKa, reaction energy (and respective uncertainties), and estimates of reaction direction from multiple independent sources. Each reaction is graded gold, silver or bronze according to the strength of the evidence behind it, so that users can weigh its reliability directly. We also release reaction directions predicted by an ensemble of large language models. This multi-source approach exposes agreements and discrepancies between sources for [~] 33,000 reactions. Finally, to ensure data integrity, a new conflict-resolution pipeline reconciles structures across sources, documenting input from curators. Our work is publicly available at https://github.com/ModelSEED/ModelSEEDDatabase.

biochemistry↗

bioRxiv2026

Structural basis for covalent inhibition of sulfatases by sulfamate warheads

The enzymatic removal of sulfate groups regulates processes ranging from steroid metabolism to carbohydrate degradation. Most sulfatases belong to the S1 family, whose members use a co-translationally installed formylglycine residue to hydrolyse sulfate esters. Arylsulfamates are potent covalent inhibitors of aryl and steroid sulfatases, including the clinical steroid sulfatase inhibitor Irosustat, yet the structure and stability of the inhibited complex remain unresolved. Arylsulfamates and carbohydrate sulfamates do not covalently inhibit many S1 carbohydrate sulfatases despite conservation of their sulfate-binding sites and formylglycine residue. Using enzyme kinetics, X-ray crystallography, molecular dynamics simulations and density functional theory calculations, we define the basis of these contrasting behaviours. High-resolution structures of the Pseudomonas aeruginosa arylsulfatase PaAtsA treated with two arylsulfamates reveal a long-lived tetrahedral, O-linked -hydroxysulfamate adduct attached to formylglycine. Molecular simulations show that replacing sulfate with sulfamate disrupts the favourable Ca2+-oxyanion interaction and alters ligand binding geometry. The permissive hydrophobic binding site of PaAtsA accommodates this rearrangement while retaining a trajectory compatible with nucleophilic attack. By contrast, in the Bacteroides thetaiotaomicron carbohydrate sulfatase BT16363S-Gal, sulfate-to-sulfamate substitution weakens binding and displaces the sulfamate from a reactive pose near the catalytic nucleophile due to a restrictive active site with conserved sugar binding. These findings define the structure and persistence of the arylsulfamate-derived covalent intermediate and explain why sulfamate warheads are tolerated by aryl sulfatases but not carbohydrate sulfatases.

biochemistry↗

bioRxiv2026

Reactivation during sleep segregates the neural representations of episodic memories

Sleep involves the reactivation of recently acquired memories, thereby shaping the neural representations supporting them. An essential feature of episodic memory (i.e., memory for events) is the link between specific elements (e.g., a cake) and the context in which they were embedded (e.g., a birthday party). We investigated how reactivation during sleep impacts item-context binding. Participants (N = 22) formed stories linking together objects in unique contexts. Using functional MRI, we measured the overlap between neural representations for contextually linked objects. During a 90-min nap, some object memories were reactivated by unobtrusively presenting object-specific sounds. Across multiple brain regions, reactivation reduced the overlap between representations of contextually linked memories, promoting object specificity. Furthermore, reactivation reduced the representational overlap between contexts, thereby promoting context specificity. Taken together, data suggest that reactivation during sleep decontextualizes memories. These results inform our understanding of how sleep contributes to interlinked memories embedded in naturalistic contexts and demonstrate that memory reactivation during sleep changes representations of memories within their distinctive contexts.

neuroscience↗

bioRxiv2025

Evaluating a structured expert elicitation approach for adaptive conservation: Lessons from five years in practice

Threatened species management relies on Ex ante estimates of species responses to different interventions to generate meaningful predictions. Structured expert elicitation is often used to generate these estimates, but comparisons of these expert-predicted outcomes with observed results are rare. This study aims to evaluate the utility of expert elicitation for adaptive management in the New South Wales Saving our Species (SoS) program in Australia by revisiting six species management plans that were generated from bespoke structured elicitation guidelines five years prior. Each species management plan included a defined scope, conceptual model, monitoring indicators and estimated response to management curves under different scenarios. Experts reviewed the conceptual models after five years of management and monitoring and compared the predicted response to management with observed monitoring data. In three of the six case studies, observed outcomes closely matched predictions. Where predictions diverged, factors such as unanticipated new threats and unexpected responses to interventions contributed to discrepancies. However, in all cases, the structured approach provided a clear logic for planning, enabling managers to systematically refine their understanding. The conceptual models and response curves proved valuable for collaboration, communication, and generating hypotheses for unexpected results. This work demonstrates the value of the bespoke guidelines in supporting adaptive management processes, strengthening the knowledge base for threatened species conservation while improving alignment between predictions and real-world outcomes.

ecology↗

bioRxiv2025

Label-free DIA-NN enabled RIME (DIANNeR) reveals glucocorticoid receptor interaction networks in breast, bladder, and blood across normaluntransformed cells, cancer cell lines, and PDXs

Understanding how transcription factors execute tissue-specific programmes requires defining their protein interaction networks in physiologically relevant contexts. However, profiling normal untransformed cells presents fundamental challenges due to intrinsically limited input material. To meet this challenge we integrated data-independent acquisition (DIA), ion mobility separation, and library-free analysis to achieve a 2-fold increase in detection without loss in enrichment. Applied to the glucocorticoid receptor (GR), a ubiquitously expressed nuclear receptor driving pleiotropic responses to standard-of-care anti-inflammatory therapeutics, DIANNeR (DIA-NN enabled RIME) resolves distinct, context-dependent networks across breast, ureter and blood in normal and transformed contexts. Key findings include: detection of a HOXA5-GR interaction in normal epithelium which was undetected in malignant models; an epithelial-restricted SMARCD3-GR interaction with prognostic relevance; and a FOXP3-BCL11B-GR network in primary CD4+ T cell populations undetected by data dependant acquisition (DDA). By defining normal GR tissue interactomes, DIANNeR provides the essential comparator for interpreting network rewiring that drives disease.

cancer biology↗

bioRxiv2025

Longitudinal TCR repertoires in ulcerative colitis patients show features distinguishing disease states

ABSTRACT AND KEY WORDSO_ST_ABSBackgroundC_ST_ABSUlcerative colitis (UC) affects an estimated 10 million people worldwide. The exact etiology of the disease is unknown, but T cell dysregulation and aberrant activation is associated with UC--prompting research into T cell receptor (TCR) repertoires in UC patients. However, few studies have compared UC patients TCR repertoires in flare (inflamed) and remission (uninflamed) states. Moreover, this is the first dataset to our knowledge examining multiple repertoires from UC patients over time, enabling longitudinal analyses. MethodsTCR repertoires were obtained for 21 patients across multiple timepoints, yielding a total of 58 samples. Repertoires clonality, diversity, overlap, and gene usage frequencies were compared across all patients. CDR3 sequences were split into K-mers (sequences of length K), and enriched K-mers in flare and remission states were identified using GLIPH2. ResultsAlthough repertoires vary across patients, there were significant differences in the usage of 20 V{beta} genes across flare and remission states. Moreover, calculating the overlap in repertoires with enriched K-mers showed higher overlap scores than when using full TCR sequences. 622 unique enriched K-mers were identified in flare states and 495 in remission states, with only 57 overlapping between the two states. ConclusionsOverall, these results highlight the importance of analyzing V{beta} gene usage and K-mer enrichment in TCR repertoires, particularly given the lack of public clones across patient cohorts. Future studies characterizing the specific antigenic targets associated with these features will pave the way for biomarker discovery in UC. KEY MESSAGESO_LIWhat is already known? UC is a chronic condition characterized by fluctuating periods of flares and remissions, as well as aberrant immune activity. C_LIO_LIWhat is new here? TCR repertoires were obtained from UC patients colonoscopies, with multiple timepoints per patient--enabling the analysis of repertoire characteristics within and across patients over time. C_LIO_LIHow can this study help patient care? Understanding the exact immune mechanisms in UC can pave the way for biomarker discovery and advance the state of care for patients, given that current treatments are non-specific and have adverse side effects. C_LI SummaryThis article presents an analysis of T cell receptor repertoires from ulcerative colitis patient samples collected during colonoscopies, identifying genes and motifs associated with patients flare (inflamed) and remission (uninflamed) states.

immunology↗

bioRxiv2025

Visible traits demonstrate that crispant founder mice can be used for phenotypic assessment

Genes can be knocked out in model organisms by introducing a single guide RNA and Cas9 into one cell zygotes. Recently, the zebrafish and Xenopus communities have employed this method in genetic screening pipelines that assess phenotypes in founders (F0), referred to as "crispants". In contrast, phenotyping crispant mice has been avoided as results are believed to be confounded by genetic mosaicism, requiring that only established mouse lines undergo phenotypic assessment. Here, we targeted seven genes associated with visible recessive phenotypes. We observed the expected null phenotype in up to 100% founders per gene. Crucially, we achieved 100% editing efficiency in all but two animals. Genetic mosaicism was common, but did not confound an animals phenotype when comprised of mutations that all disrupted the targeted gene. Mosaicism included short in-frame mutations, but these were sufficient to disrupt function of five genes. Several founders were compound heterozygotes carrying a null and a non-null allele (short in-frame mutation or late truncation), enabling functional assessment of the non-null allele to dissect protein function. Our results set the stage for using crispant founders for initial phenotypic assessment in genetic screening, before selecting candidates for further study. This will dramatically reduce animal numbers.

genetics↗

bioRxiv2025

ERM proteins regulate the shape and number of Endoplasmic Reticulum-Plasma Membrane Junctions in neurons

Endoplasmic Reticulum (ER) - Plasma Membrane (PM) Junctions (EPJs) are specialized contact sites between ER membrane and the inner leaflet of PM. These junctions are critical for lipid exchange and Ca2+ signaling. In excitable cells like neurons and muscle, EPJs further modulate membrane excitability by regulating Ca2+ homeostasis. The mechanisms controlling EPJ abundance and morphology remain poorly understood. Using in vivo fluorescence imaging and electron microscopy of C. elegans neurons, we showed that EPJs form discrete, patch-like structures distributed across the soma. Through a forward genetics screen, we identified two conserved ERM (Ezrin-Radixin-Moesin) proteins, FRM-4 and FRM-1, as key regulators of EPJ shape and abundance. Both proteins localize to EPJs and exhibit liquid-liquid phase separation properties (LLPS). in vitro, purified FRM-4 binds to FRM-1, and together bundle filamentous actin. However, their presence in LLPS condensates and actin-bundling activity are mutually exclusive. In vivo, F-actin cables surround--but do not penetrate--EPJs, where FRM proteins are enriched as phase-separated condensates. Loss of FRM-4, FRM-1, or disruption of F-actin led to increased mobility of EPJs that fused into fewer but enlarged junctions. Together, our findings demonstrate that FRM-4 and FRM-1 control EPJ morphology by organizing peri-junctional F-actin networks, thereby restricting EPJ mobility and fusion.

cell biology↗

bioRxiv2025

The BRD4-nucleosome interaction is enhanced modestly and non-selectively by histone acetylation

BRD4 regulates gene transcription in complex eukaryotes, in part through the binding of its tandem bromodomains to acetylated lysine residues found in histones and transcription factors. Despite pharmacological inhibition of these domains showing promise in preclinical studies, clinical trial data have been less encouraging so far. A stronger understanding of BRD4 biochemistry could provide a route to better outcomes. To advance on prior work, which has focused almost entirely on the binding of isolated bromodomains and acetylated peptides, we have sought the preferred nucleosomal binding partner of full-length BRD4. We demonstrate that BRD4 binds with sub-micromolar affinity to both unmodified nucleosomes and to DNA alone. In strong contrast to BRD4-peptide interactions, we also find that the affinity of BRD4 for nucleosomes is increased only 2-4-fold by histone acetylation and that this affinity has little dependence on the acetylation pattern. Despite this modest effect of acetylation, binding of BRD4 to acetyllysine in the nucleosome was more resistant to perturbation by mutation or small-molecule inhibition than BRD4-peptide interactions. Our work helps bridge the gap between cellular and prior in vitro work and provides clues to explain the in vivo chromatin occupancy profile of BRD4 and how it changes upon therapeutic inhibition. GRAPHICAL ABSTRACT O_FIG O_LINKSMALLFIG WIDTH=200 HEIGHT=79 SRC="FIGDIR/small/656505v1_ufig1.gif" ALT="Figure 1"> View larger version (22K): org.highwire.dtl.DTLVardef@1aa532aorg.highwire.dtl.DTLVardef@13a154forg.highwire.dtl.DTLVardef@4d2039org.highwire.dtl.DTLVardef@17e8377_HPS_FORMAT_FIGEXP M_FIG C_FIG

biochemistry↗

bioRxiv2025

Drug screen reveals new potent host-targeted antivirals against Mpox virus

Mpox virus (MPXV), a re-emerging zoonotic threat, has caused outbreaks in non-endemic regions through respiratory, sexual, and close-contact transmission. The increased transmissibility of Clade IIb fueled the 2022 global outbreak, with 2024 Clade Ib spread in the Democratic Republic of Congo further escalating concern. Both outbreaks were declared public health emergencies by the WHO. Although tecovirimat (TPOXX) has been used off-label for Mpox, its limited effectiveness highlights the critical need for newer antivirals for MPXV. We conducted high- throughput antiviral drug screening using a host-directed kinase inhibitor library composed of 2,750 compounds against 2022 Clade IIb MPXV. Our primary screen identified 138 compounds preventing MPXV cytopathic effects, including multiple inhibitors of EGFR, PI3K-mTOR, and Ras/Raf, as well as apoptosis and autophagy regulators. Secondary and tertiary screenings yielded a shortlist of potent, nontoxic antiviral compounds that inhibited MPXV replication. Three selected compounds, IRAK4-IN-6, SM-7368, and KRAS inhibitor-10, reduced MPXV-induced cell death in primary human epidermal keratinocytes. IRAK4-IN-6 and SM-7368 were also found to modulate NF-{kappa}B and STING signaling. Furthermore, these compounds were found effective in reducing skin lesions and viral burden in a mouse model of MPXV skin infection. Together, our study reveals new classes of antiviral compounds against MPXV, offering promising candidates for future clinical development.

microbiology↗

bioRxiv2025

Temperature and Excipient Mediated Modulation of Monoclonal Antibody Interactions Revealed by kD, Rheology, and Raman Spectroscopy

High-concentration monoclonal antibody (mAb) formulations are frequently constrained by elevated viscosity and colloidal instability, stemming from enhanced intermolecular interactions under crowded conditions. This study delineates the thermodynamic and rheological consequences of modulating protein-protein interactions through excipient-mediated and temperature-dependent mechanisms. Using an orthogonal analytical framework comprising diffusion interaction parameter (kD) measurements, high-shear rheometry, and Raman spectroscopic profiling, we interrogated mAb solutions at [~]80 and 160 mg/mL across a physiologically and industrially relevant thermal window (5-45 {degrees}C). In the absence of ionic additives, high kD values ([~]60 mL/g) indicated dominant long-range electrostatic repulsions, resulting in suppressed self-association and lower viscosity. Incorporation of NaCl (0.05% w/v) markedly decreased kD ([~]16-20 mL/g), consistent with Debye screening of surface charges and a shift toward short-range hydrophobic and van der Waals attractions, particularly impactful at elevated protein concentrations and low temperatures. Polysorbate 20 (0.05% v/v) mitigated these interactions via preferential surface adsorption, while sucrose exhibited a dualistic, concentration-dependent influence on viscosity via preferential exclusion and entropic crowding. The combination of NaCl and PS20 yielded the most pronounced rheological suppression, reflecting synergistic attenuation of both long-range repulsion and short-range association. Raman spectral analysis of Amide I/III regions confirmed structural invariance under thermal and shear stress, attributing viscosity modulation to colloidal rather than conformational perturbations. Collectively, these data elucidate the multivariate control of interparticle potentials in mAb solutions and provide a predictive basis for engineering subcutaneous formulations that optimize manufacturability, physical stability, and injectability through strategic manipulation of colloidal interaction landscapes.

biophysics↗

bioRxiv2025

Phylogenomics and macroevolution of a florally diverse Neotropical plant clade (Hillieae)

AO_SCPLOWBSTRACTC_SCPLOWHillieae is a group of [~]30 florally diverse, Neotropical epiphyte species. Species richness peaks in southern Central America and taxa display bat, hawkmoth, or hummingbird pollination syndromes. A phylogenetic framework is needed to understand floral and biogeographic evolution. We used target enrichment data to infer a species tree and a Bayesian time-calibrated tree including [~]83% of the species in the group. We inferred ancestral biogeography and pollination syndromes, described species realized bioclimatic niches via a principal component analysis, and estimated significant niche shifts using Ornstein-Uhlenbeck models to understand how different abiotic and biotic variables have shaped Hillieae evolution. We estimated that Hillieae originated in southern Central America 19 Ma and that hawkmoth pollination is the ancestral character state. Multiple independent shifts in pollination syndrome, biogeographic distribution, and realized bioclimatic niche have occurred, though bioclimatic niche is largely conserved. Using generalized linear models, we identify two interactions-- between species biogeographic ranges and pollination syndromes, and between phylogenetic covariance and pollination syndromes-- that additively affect the degree of bioclimatic niche overlap between species. Regional variation in pollination syndrome diversity and patterns of species bioclimatic niche overlap indicate a link between biogeography and species ecology in driving Hillieae diversification and syndrome evolution.

evolutionary biology↗

bioRxiv2025

A preclinical pig model of Angelman syndrome mirrors the early developmental trajectory of the human condition

Angelman syndrome is a neurodevelopmental disorder characterized by severe motor and cognitive deficits. It is caused by the loss of the maternally inherited allele of the imprinted ubiquitin-protein ligase E3A (UBE3A) gene. Rodent models of Angelman syndrome do not fully recapitulate all the symptoms associated with the condition and are limited as a preclinical model for therapeutic development. Here, we show that pigs (Sus scrofa) with a maternally inherited deletion of UBE3A (UBE3A-/+) have altered postnatal behaviors, impaired vocalizations, reduced brain growth, motor incoordination, and ataxia. Neonatal UBE3A-/+ pigs exhibited several symptoms observed in infants with Angelman syndrome, including hypotonia, suckling deficits, and failure to thrive. Collectively, these findings are consistent with the pathophysiology and developmental trajectory observed in individuals with Angelman syndrome. We anticipate that this pig model will advance our understanding of the pathophysiology of Angelman syndrome and be used as a preclinical large animal model for therapeutic development.

genetics↗

bioRxiv2025

Neuro-behavioral impact of Tourette-related striatal disinhibition in rats

Tourette syndrome has been linked to reduced GABAergic inhibition, so called neural disinhibition, in the dorsal striatum. Dorsal-striatal neural disinhibition in animal models, caused by local microinfusion of GABA-A-receptor antagonists, produces striatal local field potential (LFP) spike-wave discharges and tic-like movements resembling motor tics. Here, we characterized further the neuro-behavioral impact of striatal disinhibition, by unilateral picrotoxin infusion into the anterior dorsal striatum, in adult male Lister hooded rats. In vivo electrophysiology under anesthesia revealed enhanced neuronal burst firing in the striatum, alongside spike-wave LFP discharges. In freely moving rats, striatal picrotoxin reliably induced tic-like movements, which mainly involved lifting of the contralateral forelimb and concomitant rotational movements of head and torso, as well as occasional rotations of the whole body around its long axis. Prepulse inhibition (PPI) of the acoustic startle response was not affected, but startle reactivity tended to be reduced. Both locomotor and non-ambulatory movements, measured in photo-beam cages, were increased. Our findings suggest that, apart from generating tic-like movements, dorsal striatal disinhibition may contribute to hyperactivity, which is often comorbid with Tourette syndrome. Enhanced striatal burst firing may be important for these behavioral effects. Our findings do not support that striatal disinhibition contributes to PPI disruption, which has been associated with Tourette syndrome and suggested to contribute to tic generation. Significance StatementStriatal disinhibition has been implicated in tic generation. Using a rat model, we show that a key neural effect of striatal disinhibition is enhanced firing of striatal neurons in bursts. We also characterized further the tic-like movements caused by unilateral disinhibition of the anterior dorsal striatum. These mainly involved lifting of the contralateral forelimb and concomitant rotational movements of head and torso, as well as occasional rotations of the whole body around its long axis. Striatal disinhibition did not affect prepulse inhibition (PPI) but caused locomotor hyperactivity. This suggests that striatal disinhibition may contribute to the general hyperactivity often comorbid with Tourette syndrome, but that other brain mechanisms underlie the PPI deficits that have been reported in the condition.

neuroscience↗

bioRxiv2024

Survival time and prognostic factors in dogs diagnosed with haemangiosarcoma in UK first opinion practice

AbstractVisceral haemangiosarcoma is considered clinically aggressive in dogs, with perceived poor prognosis often leading to euthanasia at presentation. This study aimed to determine survival times and prognostic factors in dogs with haemangiosarcoma under first-opinion care. Dogs diagnosed with haemangiosarcoma in first-opinion practice in 2019 were identified in VetCompass electronic health records (EHRs) and manually examined to capture variables potentially associated with survival. Median survival time (MST) from diagnosis was calculated for the whole population and based on the primary tumour location. Binary logistic regression was used to explore differences between dogs that died on the day of diagnosis and those that survived [≥]1 day. Cox proportional hazards modelling explored factors associated with time to death in dogs that survived [≥]1 day. Across all haemangiosarcoma cases (n=788), overall MST was 9.0 days (95%CI:5.0-15.0, range: 0-1789) and proportional 1-year survival was 12.0% (95%CI:9.7-15.0%). Dogs with splenic (MST=4.0 days, 95%CI 0.0-9.0) and cutaneous haemangiosarcoma (MST=119.0 days,95%CI:85.0-248.0) had MST greater than 0 days. Male sex and increasing tumour size were associated with increased hazard of death while cutaneous location and surgery were associated with reduced hazard of death. A very short survival time was identified for haemangiosarcoma under first-opinion care. Although survival time was longest for cutaneous cases, the actualised prognosis was poor overall for haemangiosarcoma. This study provides valuable information for contextualised care and dialogues with clients in first-opinion practice.

pathology↗

bioRxiv2024

Structural determinants of inverted Alu-mediated backsplicing revealed by -MaP and -JuMP

Biogenesis of circular RNA usually involves a backsplicing reaction where the downstream donor site is ligated to the upstream acceptor site by the spliceosome. For this reaction to occur, it is hypothesized that these sites must be in proximity. Inverted repeat sequences, such as Alu elements, in the upstream and downstream introns are predicted to base-pair and represent one mechanism for inducing proximity. Here, we investigate the pre-mRNA structure of the human HIPK3 gene at exon 2, which forms a circular RNA via backsplicing. We leverage multiple chemical probing techniques, including the recently developed SHAPE- JuMP strategy, to characterize secondary and tertiary interactions in the pre- mRNA that govern backsplicing. Our data confirm that the antisense Alu elements, AluSz(-) and AluSq2(+) in the upstream and downstream introns, form a highly- paired interaction. Circularization requires formation of long-range Alu-mediated base pairs but does not require the full-length AluSq2(+). In addition to confirming long-range base pairs, our SHAPE-JuMP data identified multiple long-range interactions between non-pairing nucleotides. Genome-wide analysis of inverted repeats flanking circular RNAs confirm that their presence favors circularization, but the overall effect is modest. Together these results suggest that secondary structure considerations alone cannot fully explain backsplicing and additional interactions are key.

molecular biology↗

bioRxiv2024

Lessons learned during the journey of data: from experiment to model for predicting kinase affinity, selectivity, polypharmacology, and resistance

Recent advances in machine learning (ML) are reshaping drug discovery. Structure-based ML methods use physically-inspired models to predict binding affinities from protein:ligand complexes. These methods promise to enable the integration of data for many related targets, which addresses issues related to data scarcity for single targets and could enable generalizable predictions for a broad range of targets, including mutants. In this work, we report our experiences in building KinoML, a novel framework for ML in target-based small molecule drug discovery with an emphasis on structure-enabled methods. KinoML focuses currently on kinases as the relative structural conservation of this protein superfamily, particularly in the kinase domain, means it is possible to leverage data from the entire superfamily to make structure-informed predictions about binding affinities, selectivities, and drug resistance. Some key lessons learned in building KinoML include: the importance of reproducible data collection and deposition, the harmonization of molecular data and featurization, and the choice of the right data format to ensure reusability and reproducibility of ML models. As a result, KinoML allows users to easily achieve three tasks: accessing and curating molecular data; featurizing this data with representations suitable for ML applications; and running reproducible ML experiments that require access to ligand, protein, and assay information to predict ligand affinity. Despite KinoML focusing on kinases, this framework can be applied to other proteins. The lessons reported here can help guide the development of platforms for structure-enabled ML in other areas of drug discovery.

biophysics↗