Evolutionary dynamics of polyadenylation signals and their recognition strategies in protists.

The poly(A) signal, together with auxiliary elements, directs cleavage of a pre-mRNA and thus determines the 3' end of the mature transcript. In many species, including humans, the poly(A) signal is an AAUAAA hexamer, but we recently found that the deeply branching eukaryote Giardia lamblia uses a distinct hexamer (AGURAA) and lacks any known auxiliary elements. Our discovery prompted us to explore the evolutionary dynamics of poly(A) signals and auxiliary elements in the eukaryotic kingdom. We used direct RNA sequencing to determine poly(A) signals for four protists within the Metamonada clade (which also contains Giardia lamblia) and two outgroup protists. These experiments revealed that the AAUAAA hexamer serves as the poly(A) signal in at least four different eukaryotic clades, indicating that it is likely the ancestral signal, whereas the unusual Giardia version is derived. We found that the use and relative strengths of auxiliary elements are also surprisingly plastic; in fact, within Metamonada, species like Giardia lamblia make use of a previously unrecognized auxiliary element where nucleotides flanking the poly(A) signal itself specify genuine cleavage sites. Thus, despite the fundamental nature of pre-mRNA cleavage for the expression of all protein-coding genes, the motifs controlling this process are dynamic on evolutionary timescales, providing motivation for future biochemical and structural studies as well as new therapeutic angles to target eukaryotic pathogens.

MS treatment de-escalation: review and commentary.

Almost all currently licensed disease-modifying therapies (DMTs) for MS treatment require prolonged if not lifelong administration. Yet, as people age, the immune system has increasingly reduced responsiveness, known as immunosenescence. Many MS DMTs reduce the responsiveness of the immune system, increasing the risks for infections and possibly cancers. As people with MS (pwMS) age, it is recognized that inflammatory MS activity declines. Several studies have addressed de-escalation of DMTs for relapsing MS under special circumstances. Here, we review evidence for de-escalating DMTs as a strategy that is particularly relevant to pwMS of older age. Treatment de-escalation can involve various strategies, such as extended or reduced dosing, switching from high-efficacy DMTs having higher risks to moderately effective DMTs with lesser risks, or treatment discontinuation. Studies have suggested that for natalizumab extended dosing maintained clinical efficacy while reducing the risk of PML. Extended interval dosing of ocrelizumab mitigated the decline of Ig levels. Retrospective and observational discontinuation studies demonstrate that age is an essential modifier of drug efficacy. Discontinuation of MS treatment in older patients has been associated with a stable disease course, while younger patients who discontinued treatment were more likely to experience new clinical activity. A recently completed 2-year randomized-controlled discontinuation study in 260 stable pwMS > 55 years found stable clinical multiple sclerosis with only a small increased risk of new MRI activity upon discontinuation. DMT de-escalation or discontinuation in MS patients older than 55 years may be non-inferior to continued treatment with immunosuppressive agents having higher health risks. However, despite several small studies, a definite conclusion about treatment de-escalation in older pwMS will require larger and longer studies. Ideally, comparison of de-escalation versus continuation versus discontinuation of DMTs should be done by prospective randomized-controlled trials enrolling sufficient numbers of subjects to allow comparisons for MS patients of both sexes within age groups, such as 55-59, 60-65, 66-69, etc. Optimally, such studies should be 3 years or longer and should incorporate testing for specific markers of immunosenescence (such as T-cell receptor excision circles) to account for differential aging of individuals.

Evaluating Immune Checkpoint Blockade in Metastatic Castration-Resistant Prostate Cancers with Deleterious CDK12 Alterations in the Phase 2 IMPACT Trial.

PURPOSE: CDK12 inactivation in metastatic castration-resistant prostate cancer (mCRPC) may predict immunotherapy responses. This phase 2 trial evaluated the efficacy of immune checkpoint inhibitor (ICI) therapy in patients with CDK12-altered mCRPC. PATIENTS AND METHODS: Eligible patients had mCRPC with deleterious CDK12 alterations and any prior therapies except ICI. Cohort A received ipilimumab (1 mg/kg) with nivolumab (3 mg/kg) every 3 weeks for up to four cycles, followed by nivolumab 480 mg every 4 weeks. Cohort C received nivolumab alone 480 mg every 4 weeks. Patients with CDK12-altered nonprostate tumors were enrolled in cohort B and not reported. The primary endpoint was a 50% reduction in PSA (PSA50). Key secondary endpoints included PSA progression-free survival, overall survival, objective response rate, and safety. RESULTS: PSA was evaluable in 23 patients in cohort A and 14 in cohort C. Median lines of prior therapy were two in cohorts A and C, including any prior novel hormonal agent (74% and 79%) and chemotherapy (57% and 36%). The PSA50 rate was 9% [95% confidence interval (CI), 1%-28%] in cohort A with two responders; neither had microsatellite instability or a tumor mutational burden >10 mutations/megabase. No PSA50 responses occurred in cohort C. Median PSA progression-free survival was 7.0 months (95% CI, 3.6-11.4) in cohort A and 4.5 months (95% CI, 3.4-13.8) in cohort C. Median overall survival was 9.0 months (95% CI, 6.2-12.3) in cohort A and 13.8 months (95% CI, 3.6-not reached) in cohort C. CONCLUSIONS: There was minimal activity with ICI therapy in patients with CDK12-altered mCRPC.

Comprehensive proteogenomic characterization of rare kidney tumors.

Non-clear cell renal cell carcinomas (non-ccRCCs) encompass diverse malignant and benign tumors. Refinement of differential diagnosis biomarkers, markers for early prognosis of aggressive disease, and therapeutic targets to complement immunotherapy are current clinical needs. Multi-omics analyses of 48 non-ccRCCs compared with 103 ccRCCs reveal proteogenomic, phosphorylation, glycosylation, and metabolic aberrations in RCC subtypes. RCCs with high genome instability display overexpression of IGF2BP3 and PYCR1. Integration of single-cell and bulk transcriptome data predicts diverse cell-of-origin and clarifies RCC subtype-specific proteogenomic signatures. Expression of biomarkers MAPRE3, ADGRF5, and GPNMB differentiates renal oncocytoma from chromophobe RCC, and PIGR and SOSTDC1 distinguish papillary RCC from MTSCC. This study expands our knowledge of proteogenomic signatures, biomarkers, and potential therapeutic targets in non-ccRCC.

Discovery and Validation of a 15-Gene Prognostic Signature for Clear Cell Renal Cell Carcinoma.

PURPOSE: Develop and validate gene expression-based biomarker associated with recurrent disease to facilitate risk stratification of clear cell renal cell carcinoma (ccRCC). MATERIALS AND METHODS: We retrospectively identified 110 patients who underwent radical nephrectomy for ccRCC (discovery cohort). Patients who recurred were matched on the basis of grade/stage to patients without recurrence. Capture whole-transcriptome sequencing was performed on RNA isolated from archival tissue using the Illumina platform. We developed a gene-expression signature to predict recurrence-free survival/disease-free survival (DFS) using a 15-fold lasso and elastic-net regularized linear Cox model. We derived the 31-gene cell cycle progression (mxCCP) score using RNA-seq data for each patient. Kaplan-Meier (KM) curves and multivariable Cox proportional hazard testing were used to validate the independent prognostic impact of the gene-expression signature on DFS, disease-specific survival (DSS), and overall survival (OS) in two validation data sets (combined n = 761). RESULTS: After quality control, the discovery cohort comprised 50 patients with recurrence and 41 patients without, with a median follow-up of 26 and 36 months, respectively. We developed a 15-gene (15G) signature, which was independently associated with worse DFS and DSS (DFS: hazard ratio [HR], 11.08 [95% CI, 4.9 to 25.1]; DSS: HR, 9.67 [95% CI, 3.4 to 27.7]) in a multivariable model adjusting for clinicopathologic parameters (including stage, size, grade, and necrosis [SSIGN] score and Memorial Sloan Kettering Cancer Center nomogram) and mxCCP score. The 15G signature was also independently associated with worse DFS and DSS in both validation data sets (Validation A [n = 382], DFS: HR, 2.6 [95% CI, 1.6 to 4.3]; DSS: HR, 3 [95% CI, 1.4 to 6.1] and Validation B (n = 379), DFS: HR, 2.1 [95% CI, 1.2 to 3.6]; OS: HR, 3 [95% CI, 1.6 to 5.7]) adjusting for clinicopathologic variables and mxCCP score. CONCLUSION: We developed and validated a novel 15G prognostic signature to improve risk stratification of patients with ccRCC. Pending further validation, this signature has the potential to facilitate optimal treatment allocation.

PDGFRα signaling regulates Srsf3 transcript binding to affect PI3K signaling and endosomal trafficking.

Signaling through the platelet-derived growth factor receptor alpha (PDGFRα) plays a critical role in craniofacial development, as mutations in PDGFRA are associated with cleft lip/palate in humans and Pdgfra mutant mouse models display varying degrees of facial clefting. Phosphatidylinositol 3-kinase (PI3K)/Akt is the primary effector of PDGFRα signaling during skeletal development in the mouse. We previously demonstrated that Akt phosphorylates the RNA-binding protein serine/arginine-rich splicing factor 3 (Srsf3) downstream of PI3K-mediated PDGFRα signaling in mouse embryonic palatal mesenchyme (MEPM) cells, leading to its nuclear translocation. We further showed that ablation of Srsf3 in the murine neural crest lineage results in severe midline facial clefting, due to defects in proliferation and survival of cranial neural crest cells, and widespread alternative RNA splicing (AS) changes. Here, we sought to determine the molecular mechanisms by which Srsf3 activity is regulated downstream of PDGFRα signaling to control AS of transcripts necessary for craniofacial development. We demonstrated via enhanced UV-crosslinking and immunoprecipitation (eCLIP) of MEPM cells that PDGF-AA stimulation leads to preferential binding of Srsf3 to exons and loss of binding to canonical Srsf3 CA-rich motifs. Through the analysis of complementary RNA-seq data, we showed that Srsf3 activity results in the preferential inclusion of exons with increased GC content and lower intron to exon length ratio. Moreover, we found that the subset of transcripts that are bound by Srsf3 and undergo AS upon PDGFRα signaling commonly encode regulators of PI3K signaling and early endosomal trafficking. Functional validation studies further confirmed that Srsf3 activity downstream of PDGFRα signaling leads to retention of the receptor in early endosomes and increases in downstream PI3K-mediated Akt signaling. Taken together, our findings reveal that growth factor-mediated phosphorylation of an RNA-binding protein underlies gene expression regulation necessary for mammalian craniofacial development.

An open source knowledge graph ecosystem for the life sciences.

Translational research requires data at multiple scales of biological organization. Advancements in sequencing and multi-omics technologies have increased the availability of these data, but researchers face significant integration challenges. Knowledge graphs (KGs) are used to model complex phenomena, and methods exist to construct them automatically. However, tackling complex biomedical integration problems requires flexibility in the way knowledge is modeled. Moreover, existing KG construction methods provide robust tooling at the cost of fixed or limited choices among knowledge representation models. PheKnowLator (Phenotype Knowledge Translator) is a semantic ecosystem for automating the FAIR (Findable, Accessible, Interoperable, and Reusable) construction of ontologically grounded KGs with fully customizable knowledge representation. The ecosystem includes KG construction resources (e.g., data preparation APIs), analysis tools (e.g., SPARQL endpoint resources and abstraction algorithms), and benchmarks (e.g., prebuilt KGs). We evaluated the ecosystem by systematically comparing it to existing open-source KG construction methods and by analyzing its computational performance when used to construct 12 different large-scale KGs. With flexible knowledge representation, PheKnowLator enables fully customizable KGs without compromising performance or usability.

Estimating geographic variation of infection fatality ratios during epidemics.

Objectives: We aim to estimate geographic variability in total numbers of infections and infection fatality ratios (IFR; the number of deaths caused by an infection per 1,000 infected people) when the availability and quality of data on disease burden are limited during an epidemic. Methods: We develop a noncentral hypergeometric framework that accounts for differential probabilities of positive tests and reflects the fact that symptomatic people are more likely to seek testing. We demonstrate the robustness, accuracy, and precision of this framework, and apply it to the United States (U.S.) COVID-19 pandemic to estimate county-level SARS-CoV-2 IFRs. Results: The estimators for the numbers of infections and IFRs showed high accuracy and precision; for instance, when applied to simulated validation data sets, across counties, Pearson correlation coefficients between estimator means and true values were 0.996 and 0.928, respectively, and they showed strong robustness to model misspecification. Applying the county-level estimators to the real, unsimulated COVID-19 data spanning April 1, 2020 to September 30, 2020 from across the U.S., we found that IFRs varied from 0 to 44.69, with a standard deviation of 3.55 and a median of 2.14. Conclusions: The proposed estimation framework can be used to identify geographic variation in IFRs across settings.

Integrating biological knowledge for mechanistic inference in the host-associated microbiome.

Advances in high-throughput technologies have enhanced our ability to describe microbial communities as they relate to human health and disease. Alongside the growth in sequencing data has come an influx of resources that synthesize knowledge surrounding microbial traits, functions, and metabolic potential with knowledge of how they may impact host pathways to influence disease phenotypes. These knowledge bases can enable the development of mechanistic explanations that may underlie correlations detected between microbial communities and disease. In this review, we survey existing resources and methodologies for the computational integration of broad classes of microbial and host knowledge. We evaluate these knowledge bases in their access methods, content, and source characteristics. We discuss challenges of the creation and utilization of knowledge bases including inconsistency of nomenclature assignment of taxa and metabolites across sources, whether the biological entities represented are rooted in ontologies or taxonomies, and how the structure and accessibility limit the diversity of applications and user types. We make this information available in a code and data repository at: https://github.com/lozuponelab/knowledge-source-mappings. Addressing these challenges will allow for the development of more effective tools for drawing from abundant knowledge to find new insights into microbial mechanisms in disease by fostering a systematic and unbiased exploration of existing information.

Patient and Parent Experience with Pediatric Care Providers During the COVID-19 Pandemic: A Comparison of Press Ganey Survey Scores for Telehealth and In-Person Encounters.

Background: Prior research suggests that pediatric patients and their parents/guardians are generally satisfied with care provided through telehealth. The objective of this study was to compare Press Ganey provider-oriented experience survey scores between telehealth and in-person patient encounters among a variety of pediatric clinical specialties at a large academic medical center. Methods: We analyzed Press Ganey survey data from pediatric patient encounters from UC Davis Health, collected between August 2020 and February 2022. Survey results analyzed respondents' satisfaction with care providers, including satisfaction with explanations given, discussions led, concern showed, and inclusion by providers; and the likelihood the survey respondent would recommend the provider to others. We used logistic regression models, which included case mix variables and clinical specialty to compare the odds of scoring the highest possible survey response ("top box" score). Results: Of the 6,093 survey responses that met inclusion criteria, 1,157 (19%) were associated with telehealth encounters and 4,936 (81%) were associated with in-person encounters. We found no significant difference in the odds of respondents giving a top box score to rate their satisfaction with their care provider between telehealth and in-person encounters. When respondents were asked whether they would recommend the care provider to others, the odds of giving a top box score following a telehealth encounter relative to an in-person encounter was 1.22 (95% confidence interval [0.97-1.52]; p-value = 0.09). Discussion: We found that survey respondents' experiences with their care provider are high and comparable for telehealth and in-person encounters in a pediatric population.

Intrathecal administration of mesenchymal stem cells in patients with adrenomyeloneuropathy.

Background and objectives: X-linked adrenomyeloneuropathy (AMN) is an inherited neurodegenerative disorder associated with mutations in the ABCD1 gene and the accumulation of very long-chain fatty acids (VLFCAs) in plasma and tissues. Currently, there is no effective treatment for AMN. We have aimed to evaluate the therapeutic effects of mesenchymal stem cell (MSC) transplantation in patients with AMN. Methods: This is a small cohort open-label study with patients with AMN diagnosed and treated at the University Hospital in Olsztyn, Poland. All patients met clinical, biochemical, MRI, and neuropsychological criteria for AMN. MSCs derived from Wharton jelly, 20 × 106 cells, were administered intrathecally three times every 2 months, and patients were followed up for an additional 3 months. The primary outcome measures included a blinded assessment of lower limb muscle strength with the Medical Research Council Manual Muscle Testing scale at baseline and on every month visits until the end of the study. Additional outcomes included measurements of the timed 25-feet walk (T25FW) and VLFCA serum ratio. Results: Three male patients with AMN with an age range of 26-37 years participated in this study. All patients experienced increased muscle strength in the lower limbs at the end of the study versus baseline. The power grade increased by 25-43% at the baseline. In addition, all patients showed an improvement trend in walking speed measured with the T25FW test. Treatment with MSCs in patients with AMN appeared to be safe and well tolerated. Discussion: The results of this study demonstrated that intrathecal administration of WJ-MSC improves motor symptoms in patients with AMN. The current findings lend support to the safety and feasibility of MSC therapy as a potentially viable treatment option for patients with AMN.

Structured Prompt Interrogation and Recursive Extraction of Semantics (SPIRES): a method for populating knowledge bases using zero-shot learning.

MOTIVATION: Creating knowledge bases and ontologies is a time consuming task that relies on manual curation. AI/NLP approaches can assist expert curators in populating these knowledge bases, but current approaches rely on extensive training data, and are not able to populate arbitrarily complex nested knowledge schemas. RESULTS: Here we present Structured Prompt Interrogation and Recursive Extraction of Semantics (SPIRES), a Knowledge Extraction approach that relies on the ability of Large Language Models (LLMs) to perform zero-shot learning and general-purpose query answering from flexible prompts and return information conforming to a specified schema. Given a detailed, user-defined knowledge schema and an input text, SPIRES recursively performs prompt interrogation against an LLM to obtain a set of responses matching the provided schema. SPIRES uses existing ontologies and vocabularies to provide identifiers for matched elements. We present examples of applying SPIRES in different domains, including extraction of food recipes, multi-species cellular signaling pathways, disease treatments, multi-step drug mechanisms, and chemical to disease relationships. Current SPIRES accuracy is comparable to the mid-range of existing Relation Extraction methods, but greatly surpasses an LLM's native capability of grounding entities with unique identifiers. SPIRES has the advantage of easy customization, flexibility, and, crucially, the ability to perform new tasks in the absence of any new training data. This method supports a general strategy of leveraging the language interpreting capabilities of LLMs to assemble knowledge bases, assisting manual knowledge curation and acquisition while supporting validation with publicly-available databases and ontologies external to the LLM. AVAILABILITY AND IMPLEMENTATION: SPIRES is available as part of the open source OntoGPT package: https://github.com/monarch-initiative/ontogpt.

Norepinephrine induces anoikis resistance in high-grade serous ovarian cancer precursor cells.

High-grade serous carcinoma (HGSC) is the most lethal gynecological malignancy in the United States. Late diagnosis and the emergence of chemoresistance have prompted studies into how the tumor microenvironment, and more recently tumor innervation, may be leveraged for HGSC prevention and interception. In addition to stess-induced sources, concentrations of the sympathetic neurotransmitter norepinephrine (NE) in the ovary increase during ovulation and after menopause. Importantly, NE exacerbates advanced HGSC progression. However, little is known about the role of NE in early disease pathogenesis. Here, we investigated the role of NE in instigating anchorage independence and micrometastasis of preneoplastic lesions from the fallopian tube epithelium (FTE) to the ovary, an essential step in HGSC onset. We found that in the presence of NE, FTE cell lines were able to survive in ultra-low-attachment (ULA) culture in a ß-adrenergic receptor-dependent (ß-AR-dependent) manner. Importantly, spheroid formation and cell viability conferred by treatment with physiological sources of NE were abrogated using the ß-AR blocker propranolol. We have also identified that NE-mediated anoikis resistance may be attributable to downregulation of colony-stimulating factor 2. These findings provide mechanistic insight and identify targets that may be regulated by ovary-derived NE in early HGSC.

Systematic analysis of nonsense variants uncovers peptide release rate as a novel modifier of nonsense-mediated mRNA decay efficiency.

Nonsense variants underlie many genetic diseases. The phenotypic impact of nonsense variants is determined by Nonsense-mediated mRNA decay (NMD), which degrades transcripts with premature termination codons (PTCs). NMD activity varies across transcripts and cellular contexts via poorly understood mechanisms. Here, by leveraging human genetic datasets, we uncover that the amino acid preceding the PTC dramatically affects NMD activity in human cells. We find that glycine codons in particular support high levels of NMD and are enriched before PTCs but depleted before normal termination codons (NTCs). Gly-PTC enrichment is most pronounced in human genes that tolerate loss-of-function variants. This suggests a strong biological impact for Gly-PTC in ensuring robust elimination of potentially toxic truncated proteins from non-essential genes. Biochemical assays revealed that the peptide release rate during translation termination is highly dependent on the identity of the amino acid preceding the stop codon. This release rate is the most critical feature determining NMD activity across our massively parallel reporter assays. Together, we conclude that NMD activity is significantly modulated by the "window of opportunity" offered by translation termination kinetics. Integrating the window of opportunity model with the existing framework of NMD would enable more accurate nonsense variant interpretation in the clinic.

A bacterial sensor taxonomy across earth ecosystems for machine learning applications.

Microbial communities have evolved to colonize all ecosystems of the planet, from the deep sea to the human gut. Microbes survive by sensing, responding, and adapting to immediate environmental cues. This process is driven by signal transduction proteins such as histidine kinases, which use their sensing domains to bind or otherwise detect environmental cues and "transduce" signals to adjust internal processes. We hypothesized that an ecosystem's unique stimuli leave a sensor "fingerprint," able to identify and shed insight on ecosystem conditions. To test this, we collected 20,712 publicly available metagenomes from Host-associated, Environmental, and Engineered ecosystems across the globe. We extracted and clustered the collection's nearly 18M unique sensory domains into 113,712 similar groupings with MMseqs2. We built gradient-boosted decision tree machine learning models and found we could classify the ecosystem type (accuracy: 87%) and predict the levels of different physical parameters (R2 score: 83%) using the sensor cluster abundance as features. Feature importance enables identification of the most predictive sensors to differentiate between ecosystems which can lead to mechanistic interpretations if the sensor domains are well annotated. To demonstrate this, a machine learning model was trained to predict patient's disease state and used to identify domains related to oxygen sensing present in a healthy gut but missing in patients with abnormal conditions. Moreover, since 98.7% of identified sensor domains are uncharacterized, importance ranking can be used to prioritize sensors to determine what ecosystem function they may be sensing. Furthermore, these new predictive sensors can function as targets for novel sensor engineering with applications in biotechnology, ecosystem maintenance, and medicine.IMPORTANCEMicrobes infect, colonize, and proliferate due to their ability to sense and respond quickly to their surroundings. In this research, we extract the sensory proteins from a diverse range of environmental, engineered, and host-associated metagenomes. We trained machine learning classifiers using sensors as features such that it is possible to predict the ecosystem for a metagenome from its sensor profile. We use the optimized model's feature importance to identify the most impactful and predictive sensors in different environments. We next use the sensor profile from human gut metagenomes to classify their disease states and explore which sensors can explain differences between diseases. The sensors most predictive of environmental labels here, most of which correspond to uncharacterized proteins, are a useful starting point for the discovery of important environment signals and the development of possible diagnostic interventions.

Linking Mechanisms of Vitamin D Signaling with Multiple Sclerosis.

Environmental triggers often work via signal transduction cascades that modulate the epigenome and transcriptome of cell types involved in the disease process. Multiple sclerosis (MS) is an autoimmune disease affecting the central nervous system being characterized by a combination of recurring inflammation, demyelination and progressive loss of axons. The mechanisms of MS onset are not fully understood and genetic variants may explain only some 20% of the disease susceptibility. From the environmental factors being involved in disease development low vitamin D levels have been shown to significantly contribute to MS susceptibility. The pro-hormone vitamin D3 acts via its metabolite 1α,25-dihydroxyvitamin D3 (1,25(OH)2D3) as a high affinity ligand to the transcription factor VDR (vitamin D receptor) and is a potent modulator of the epigenome at thousands of genomic regions and the transcriptome of hundreds of genes. A major target tissue of the effects of 1,25(OH)2D3 and VDR are cells of innate and adaptive immunity, such as monocytes, dendritic cells as well as B and T cells. Vitamin D induces immunological tolerance in T cells and reduces inflammatory reactions of various types of immune cells, all of which are implicated in MS pathogenesis. The immunomodulatory effects of 1,25(OH)2D3 contribute to the prevention of MS. However, the strength of the responses to vitamin D3 supplementation is highly variegated between individuals. This review will relate mechanisms of individual's vitamin D responsiveness to MS susceptibility and discuss the prospect of vitamin D3 supplementation as a way to extinguish the autoimmunity in MS.

Gene Set Summarization using Large Language Models.

Molecular biologists frequently interpret gene lists derived from high-throughput experiments and computational analysis. This is typically done as a statistical enrichment analysis that measures the over- or under-representation of biological function terms associated with genes or their properties, based on curated assertions from a knowledge base (KB) such as the Gene Ontology (GO). Interpreting gene lists can also be framed as a textual summarization task, enabling the use of Large Language Models (LLMs), potentially utilizing scientific texts directly and avoiding reliance on a KB. We developed SPINDOCTOR (Structured Prompt Interpolation of Natural Language Descriptions of Controlled Terms for Ontology Reporting), a method that uses GPT models to perform gene set function summarization as a complement to standard enrichment analysis. This method can use different sources of gene functional information: (1) structured text derived from curated ontological KB annotations, (2) ontology-free narrative gene summaries, or (3) direct model retrieval. We demonstrate that these methods are able to generate plausible and biologically valid summary GO term lists for gene sets. However, GPT-based approaches are unable to deliver reliable scores or p-values and often return terms that are not statistically significant. Crucially, these methods were rarely able to recapitulate the most precise and informative term from standard enrichment, likely due to an inability to generalize and reason using an ontology. Results are highly nondeterministic, with minor variations in prompt resulting in radically different term lists. Our results show that at this point, LLM-based methods are unsuitable as a replacement for standard term enrichment analysis and that manual curation of ontological assertions remains necessary.

KG-Hub-building and exchanging biological knowledge graphs.

MOTIVATION: Knowledge graphs (KGs) are a powerful approach for integrating heterogeneous data and making inferences in biology and many other domains, but a coherent solution for constructing, exchanging, and facilitating the downstream use of KGs is lacking. RESULTS: Here we present KG-Hub, a platform that enables standardized construction, exchange, and reuse of KGs. Features include a simple, modular extract-transform-load pattern for producing graphs compliant with Biolink Model (a high-level data model for standardizing biological data), easy integration of any OBO (Open Biological and Biomedical Ontologies) ontology, cached downloads of upstream data sources, versioned and automatically updated builds with stable URLs, web-browsable storage of KG artifacts on cloud infrastructure, and easy reuse of transformed subgraphs across projects. Current KG-Hub projects span use cases including COVID-19 research, drug repurposing, microbial-environmental interactions, and rare disease research. KG-Hub is equipped with tooling to easily analyze and manipulate KGs. KG-Hub is also tightly integrated with graph machine learning (ML) tools which allow automated graph ML, including node embeddings and training of models for link prediction and node classification. AVAILABILITY AND IMPLEMENTATION: https://kghub.org.