An Interpretable Machine Learning Framework for Rare Disease: A Case Study to Stratify Infection Risk in Pediatric Leukemia.

Background: Datasets on rare diseases, like pediatric acute myeloid leukemia (AML) and acute lymphoblastic leukemia (ALL), have small sample sizes that hinder machine learning (ML). The objective was to develop an interpretable ML framework to elucidate actionable insights from small tabular rare disease datasets. Methods: The comprehensive framework employed optimized data imputation and sampling, supervised and unsupervised learning, and literature-based discovery (LBD). The framework was deployed to assess treatment-related infection in pediatric AML and ALL. Results: An interpretable decision tree classified the risk of infection as either "high risk" or "low risk" in pediatric ALL (n = 580) and AML (n = 132) with accuracy of ∼79%. Interpretable regression models predicted the discrete number of developed infections with a mean absolute error (MAE) of 2.26 for bacterial infections and an MAE of 1.29 for viral infections. Features that best explained the development of infection were the chemotherapy regimen, cancer cells in the central nervous system at initial diagnosis, chemotherapy course, leukemia type, Down syndrome, race, and National Cancer Institute risk classification. Finally, SemNet 2.0, an open-source LBD software that links relationships from 33+ million PubMed articles, identified additional features for the prediction of infection, like glucose, iron, neutropenia-reducing growth factors, and systemic lupus erythematosus (SLE). Conclusions: The developed ML framework enabled state-of-the-art, interpretable predictions using rare disease tabular datasets. ML model performance baselines were successfully produced to predict infection in pediatric AML and ALL.

Catalytic Light-Driven Strategy for Transforming Oximes to Carbonyls.

Oxime and carbonyl functional groups serve as powerful chemical hubs for constructing complex synthetic targets and valuable molecular scaffolds. In furthering this value, we report a photopromoted catalytic deoximation protocol for converting oximes and their derivatives to carbonyl functional groups. This strategic approach benefits from the use of renewable light energy input and ambient air conditions, in addition to demonstrating good substrate scope, functional group tolerance, and product yields. In offering, insights into these reactivity mechanistic studies are communicated, and the value of this protocol is further shown through one-pot operations.

A Boron Scan of Ethyl Acetoacetate Leads to Versatile Building Blocks.

Discovered in the 19th century, ethyl acetoacetate has been central to the development of organic chemistry, including its pedagogy and applications. In this study, we present borylated derivatives of this venerable molecule. A boron handle has been installed at either α ${{\rm \alpha }}$ - or ß ${\beta }$ -position of acetoacetate by homologation of acyl-MIDA (N-methyliminodiacetic acid) boronates with diazoacetates. Either alkyl or boryl groups were found to migrate with regiochemistry being a function of the steric bulk of the diazo species. Boryl ß ${{\rm \beta }}$ -ketoesters can be further modified into borylated pyrazolones and oximes, thereby expanding the synthetic toolkit and offering opportunities for additional modifications.

Ionizable Lipid Nanoparticles for Therapeutic Base Editing of Congenital Brain Disease.

Delivery of mRNA-based therapeutics to the perinatal brain holds great potential in treating congenital brain diseases. However, nonviral delivery platforms that facilitate nucleic acid delivery in this environment have yet to be rigorously studied. Here, we screen a diverse library of ionizable lipid nanoparticles (LNPs) via intracerebroventricular (ICV) injection in both fetal and neonatal mice and identify an LNP formulation with greater functional mRNA delivery in the perinatal brain than an FDA-approved industry standard LNP. Following in vitro optimization of the top-performing LNP (C3 LNP) for codelivery of an adenine base editing platform, we improve the biochemical phenotype of a lysosomal storage disease in the neonatal mouse brain, exhibit proof-of-principle mRNA brain transfection in vivo in a fetal nonhuman primate model, and demonstrate the translational potential of C3 LNPs ex vivo in human patient-derived brain tissues. These LNPs may provide a clinically translatable platform for in utero and postnatal mRNA therapies including gene editing in the brain.

Fetal allotransplant recipients are resistant to graft-versus-host disease.

In utero hematopoietic cell transplantation (IUHCT) is an experimental treatment for congenital hemoglobinopathies, including Sickle cell disease and thalassemias. One of the principal advantages of IUHCT is the predisposition of the developing fetus toward immunologic tolerance. This allows for engraftment across immune barriers without immunosuppression and, potentially, decreased susceptibility to graft-versus-host disease (GVHD). We demonstrate fetal resistance to GVHD following T cell-replete allogeneic hematopoietic cell transplantation compared with the neonate. We show that this resistance is associated with elevated fetal serum interleukin-10 conducive to the induction of regulatory T cells (Tregs). Finally, we demonstrate that the adoptive transfer of Tregs from IUHCT recipients to neonates uniformly prevents GVHD, recapitulating the predisposition to tolerance observed after fetal allotransplantation. These findings demonstrate fetal resistance to GVHD following hematopoietic cell transplantation and elucidate Tregs as important contributors.

Enteral nutrition support for infants with pulmonary hypoplasia: A qualitative evaluation of caregiver and provider perspectives.

BACKGROUND: Enteral nutrition is a critical intervention that supports the growth of children with pulmonary hypoplasia (PH). We explored the experiences of caregivers and providers caring for children with PH to better understand gaps in knowledge transfer and identify barriers and facilitators to caregiving to inform interventions that may improve support. METHODS: This qualitative study included 10 interviews with caregivers and 10 clinical team members at a single integrated care program for children with PH. An inductive and iterative coding strategy was employed to produce a codebook. After cluster analysis, themes were generated to capture participant sentiments. RESULTS: Themes were defined along a care continuum (1) initiation, (2) adaptation, and (3) maintenance that represented distinct phases of adjustment to enteral nutrition support (1) in the perinatal period and initial neonatal intensive care unit (NICU) admission, (2) from discharge planning through the family's first days at home and establishment of a stable feeding regime, and (3) through long-term follow-up and weaning. Notable subthemes included uncertainty, partnerships in training, and obstacles to adaptation. CONCLUSIONS: Among children with PH, the caregiver-provider relationship during the perinatal and NICU course is critical to promoting caregiver adaptation to the needs of the child. Ongoing considerations to support resource alignment and transition to a stable feeding regimen may facilitate caregiver adjustment to a "new normal," culminating in successful growth and/or weaning. These findings will inform interventions focused on training curricula, discharge planning, and the provision of follow-up in the context of an integrated care program for PH.

Amniotic fluid stabilized lipid nanoparticles for in utero intra-amniotic mRNA delivery.

Congenital disorders resulting in pathological protein deficiencies are most often treated postnatally with protein or enzyme replacement therapies. However, treatment of these disorders in utero before irreversible disease onset could significantly minimize disease burden, morbidity, and mortality. One possible strategy for the prenatal treatment of congenital disorders is in utero delivery of messenger RNA (mRNA). mRNA is a nucleic acid therapeutic that has previously been investigated as a platform for protein replacement therapies and gene editing technologies. While viral vectors have been explored to induce intracellular expression of mRNA, they are limited in their clinical application due to risks associated with immunogenicity and genomic integration. As an alternative to viral vectors, safe and efficient in utero mRNA delivery can be achieved using ionizable lipid nanoparticles (LNPs). While LNPs have demonstrated potent in vivo mRNA delivery to the liver following intravenous administration, intra-amniotic delivery has the potential to deliver mRNA to cells and tissues beyond those in the liver, such as in the skin, lung, and digestive tract. However, LNP stability in fetal amniotic fluid and how this stability affects mRNA delivery has not been previously investigated. Here, we engineered a library of LNPs using orthogonal design of experiments (DOE) to evaluate how LNP structure affects their stability in amniotic fluid ex utero and whether a lead candidate identified from these stability measurements enables intra-amniotic mRNA delivery in utero. We used a combination of techniques including dynamic light scattering (DLS), transmission electron microscopy (TEM), and chromatography followed by protein content quantification to screen LNP stability in amniotic fluids. These results identified multiple lead LNP formulations that are highly stable in amniotic fluids ranging from small animals to humans, including mouse, sheep, pig, and human amniotic fluid samples. We then demonstrate that stable LNPs from the ex utero screen in mouse amniotic fluid enabled potent mRNA delivery in primary fetal lung fibroblasts and in utero following intra-amniotic injection in a murine model. This exploration of ex utero stability in amniotic fluids demonstrates a means by which to identify novel LNP formulations for prenatal treatment of congenital disorders via in utero mRNA delivery.

Generation and reactivity of an elusive base-stabilised phosphinidene.

Reduction of phosphorus dichloride 6, supported by the diaryloxyphenyl group (OCO) featuring two bulky phenoxy wingtips, by PMe3, generates a reactive intermediate that behaves as a base-stabilized phosphinidene (OCO)P (5). Warming up a solution of this species in toluene to room temperature results in trimerization to give the isolable cyclic triphosphine [(OCO)P]3, whereas in situ trapping with 2,3-dimethylbutadiene-1,3 afforded a 3,4-dimethylphospholene-3. Investigation of the reduction of 6 by the phosphine PMe3 by NMR led to the observation of a persistent species between -10 °C and 10 °C. A DFT study of this process suggests that this compound cannot be the proposed phosphinidene 5, and is more likely the diphosphine (OCO)ClP-PCl(OCO) (12). Attempted reduction of 5 by the bulky carbene IPr resulted in unusual electrophilic substitution in the carbene olefin backbone by the chlorophosphinyl group.

Mandible Biomechanics and Continuously Erupting Teeth: A New Defect Model for Studying Load-Bearing Biomaterials.

Animals with elodont dentition and unfused mandible symphyses are hypothesized to have symmetric incisor morphology. Since these animals maintain their teeth by gnawing, they may provide physiologic feedback on mechanical function when unilateral mandible defects are created that manifest as ipsilateral changes in tooth structure. This defect model would potentially generate important information on the functional/mechanical properties of implants. Rats' and rabbits' mandibles and teeth are analyzed with µCT at baseline and post-intervention (n = 8 for each). Baseline incisors were compared. In a unilateral mandible pilot study, defects-ranging from critical size defect to complete ramus osteotomies-were created to assess effect on dentition (rats, n = 7; rabbits, n = 6). Within 90% confidence intervals, animals showed no baseline left/right differences in their incisors. There are apparent dental changes associated with unilateral defect type and location. Thus, at baseline, animals exhibit statistically significant incisor symmetry and there is an apparent relationship between mandible defect and incisor growth. The baseline symmetry proven here sets the stage to study the degree to which hemi-mandible destabilizing procedures result in measurable & reproducible disruption of dental asymmetry. In a validated model, an implant designed to function under load that prevents incisor asymmetry would provide supporting evidence that the implant has clinically useful load-bearing function.

In utero adenine base editing corrects multi-organ pathology in a lethal lysosomal storage disease.

In utero base editing has the potential to correct disease-causing mutations before the onset of pathology. Mucopolysaccharidosis type I (MPS-IH, Hurler syndrome) is a lysosomal storage disease (LSD) affecting multiple organs, often leading to early postnatal cardiopulmonary demise. We assessed in utero adeno-associated virus serotype 9 (AAV9) delivery of an adenine base editor (ABE) targeting the Idua G→A (W392X) mutation in the MPS-IH mouse, corresponding to the common IDUA G→A (W402X) mutation in MPS-IH patients. Here we show efficient long-term W392X correction in hepatocytes and cardiomyocytes and low-level editing in the brain. In utero editing was associated with improved survival and amelioration of metabolic, musculoskeletal, and cardiac disease. This proof-of-concept study demonstrates the possibility of efficiently performing therapeutic base editing in multiple organs before birth via a clinically relevant delivery mechanism, highlighting the potential of this approach for MPS-IH and other genetic diseases.

Effects of a Bioprocessed Soybean Meal Ingredient on the Intestinal Microbiota of Hybrid Striped Bass, Morone chrysops x M. saxatilis.

The hybrid striped bass (Morone chrysops x M. saxatilis) is a carnivorous species and a major product of US aquaculture. To reduce costs and improve resource sustainability, traditional ingredients used in fish diets are becoming more broadly replaced by plant-based products; however, plant meals can be problematic for carnivorous fish. Bioprocessing has improved nutritional quality and allowed higher inclusions in fish diets, but these could potentially affect other systems such as the gut microbiome. In this context, the effects of bioprocessed soybean meal on the intestinal bacterial composition in hybrid striped bass were investigated. Using high-throughput sequencing of amplicons targeting the V1-V3 region of the 16S rRNA gene, no significant difference in bacterial composition was observed between fish fed a control diet, and fish fed a diet with the base bioprocessed soybean meal. The prominent Operational Taxonomic Unit (OTU) in these samples was predicted to be a novel species affiliated to Peptostreptococcaceae. In contrast, the intestinal bacterial communities of fish fed bioprocessed soybean meal that had been further modified after fermentation exhibited lower alpha diversity (p < 0.05), as well as distinct and more varied composition patterns, with OTUs predicted to be strains of Lactococcus lactis, Plesiomonas shigelloides, or Ralstonia pickettii being the most dominant. Together, these results suggest that compounds in bioprocessed soybean meal can affect intestinal bacterial communities in hybrid striped bass.

Biomedical Text Link Prediction for Drug Discovery: A Case Study with COVID-19.

Link prediction in artificial intelligence is used to identify missing links or derive future relationships that can occur in complex networks. A link prediction model was developed using the complex heterogeneous biomedical knowledge graph, SemNet, to predict missing links in biomedical literature for drug discovery. A web application visualized knowledge graph embeddings and link prediction results using TransE, CompleX, and RotatE based methods. The link prediction model achieved up to 0.44 hits@10 on the entity prediction tasks. The recent outbreak of severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2), also known as COVID-19, served as a case study to demonstrate the efficacy of link prediction modeling for drug discovery. The link prediction algorithm guided identification and ranking of repurposed drug candidates for SARS-CoV-2 primarily by text mining biomedical literature from previous coronaviruses, including SARS and middle east respiratory syndrome (MERS). Repurposed drugs included potential primary SARS-CoV-2 treatment, adjunctive therapies, or therapeutics to treat side effects. The link prediction accuracy for nodes ranked highly for SARS coronavirus was 0.875 as calculated by human in the loop validation on existing COVID-19 specific data sets. Drug classes predicted as highly ranked include anti-inflammatory, nucleoside analogs, protease inhibitors, antimalarials, envelope proteins, and glycoproteins. Examples of highly ranked predicted links to SARS-CoV-2: human leukocyte interferon, recombinant interferon-gamma, cyclosporine, antiviral therapy, zidovudine, chloroquine, vaccination, methotrexate, artemisinin, alkaloids, glycyrrhizic acid, quinine, flavonoids, amprenavir, suramin, complement system proteins, fluoroquinolones, bone marrow transplantation, albuterol, ciprofloxacin, quinolone antibacterial agents, and hydroxymethylglutaryl-CoA reductase inhibitors. Approximately 40% of identified drugs were not previously connected to SARS, such as edetic acid or biotin. In summary, link prediction can effectively suggest repurposed drugs for emergent diseases.

Meta-Analysis of Gastrointestinal Adverse Events from Tyrosine Kinase Inhibitors for Chronic Myeloid Leukemia.

Tyrosine kinase inhibitors (TKIs) are the frontline therapy for BCR-ABL (Ph+) chronic myeloid leukemia (CML). A systematic meta-analysis of 43 peer-reviewed studies with 10,769 CML patients compared the incidence of gastrointestinal adverse events (GI AEs) in a large heterogeneous CML population as a function of TKI type. Incidence and severity of nausea, vomiting, and diarrhea were assessed for imatinib, dasatinib, bosutinib, and nilotinib. Examination of combined TKI average GI AE incidence found diarrhea most prevalent (22.5%), followed by nausea (20.6%), and vomiting (12.9%). Other TKI GI AEs included constipation (9.2%), abdominal pain (7.6%), gastrointestinal hemorrhage (3.5%), and pancreatitis (2.2%). Mean GI AE incidence was significantly different between TKIs (p < 0.001): bosutinib (52.9%), imatinib (24.2%), dasatinib (20.4%), and nilotinib (9.1%). Diarrhea was the most prevalent GI AE with bosutinib (79.2%) and dasatinib (28.1%), whereas nausea was most prevalent with imatinib (33.0%) and nilotinib (13.2%). Incidence of grade 3 or 4 severe GI AEs was ≤3% except severe diarrhea with bosutinib (9.5%). Unsupervised clustering revealed treatment efficacy measured by the complete cytogenetic response, major molecular response, and overall survival is driven most by disease severity, not TKI type. For patients with chronic phase CML without resistance, optimal TKI selection should consider TKI AE profile, comorbidities, and lifestyle.

Ionizable lipid nanoparticles for in utero mRNA delivery.

Clinical advances enable the prenatal diagnosis of genetic diseases that are candidates for gene and enzyme therapies such as messenger RNA (mRNA)-mediated protein replacement. Prenatal mRNA therapies can treat disease before the onset of irreversible pathology with high therapeutic efficacy and safety due to the small fetal size, immature immune system, and abundance of progenitor cells. However, the development of nonviral platforms for prenatal delivery is nascent. We developed a library of ionizable lipid nanoparticles (LNPs) for in utero mRNA delivery to mouse fetuses. We screened LNPs for luciferase mRNA delivery and identified formulations that accumulate within fetal livers, lungs, and intestines with higher efficiency and safety compared to benchmark delivery systems, DLin-MC3-DMA and jetPEI. We demonstrate that LNPs can deliver mRNAs to induce hepatic production of therapeutic secreted proteins. These LNPs may provide a platform for in utero mRNA delivery for protein replacement and gene editing.

Molecular Differentiation of Four Species of Oropsylla (Siphonaptera: Ceratophyllidae) Using PCR-Based Single Strand Conformation Polymorphism Analyses and DNA Sequencing.

It is often difficult to distinguish morphologically between closely related species of fleas (Siphonaptera). Morphological identification of fleas often requires microscopic examination of internal structures in specimens cleared using caustic solutions. This process degrades DNA and/or inhibits DNA extraction from specimens, which limits molecular-based studies on individual fleas and their microbiomes. Our objective was to distinguish between Oropsylla rupestris (Jordan), Oropsylla tuberculata (Baker), Oropsylla bruneri (Baker), and Oropsylla labis (Jordan & Rothschild) (Ceratophyllidae) using PCR-based single strand conformation polymorphism (SSCP) analyses and DNA sequencing. A 446 bp region of the nuclear 28S ribosomal RNA (rRNA) gene was used as the genetic marker. The results obtained for 36 reference specimens (i.e., fleas that were morphologically identified to species) revealed no intraspecific variation in DNA sequence, whereas the DNA sequences of the four species of Oropsylla differed from one another at two to six nucleotide positions. Each flea species also had a unique SSCP banding pattern. SSCP analyses were then used to identify another 84 fleas that had not been identified morphologically. DNA sequencing data confirmed the species identity of fleas subjected to SSCP. This demonstrates that PCR-SSCP combined with DNA sequencing of the 28S rRNA gene is a very effective approach for the delineation of four closely related species of flea.

Publisher Correction: Inference of transcription factor binding from cell-free DNA enables tumor subtype prediction and early detection.

An amendment to this paper has been published and can be accessed via a link at the top of the paper.

Inference of transcription factor binding from cell-free DNA enables tumor subtype prediction and early detection.

Deregulation of transcription factors (TFs) is an important driver of tumorigenesis, but non-invasive assays for assessing transcription factor activity are lacking. Here we develop and validate a minimally invasive method for assessing TF activity based on cell-free DNA sequencing and nucleosome footprint analysis. We analyze whole genome sequencing data for >1,000 cell-free DNA samples from cancer patients and healthy controls using a bioinformatics pipeline developed by us that infers accessibility of TF binding sites from cell-free DNA fragmentation patterns. We observe patient-specific as well as tumor-specific patterns, including accurate prediction of tumor subtypes in prostate cancer, with important clinical implications for the management of patients. Furthermore, we show that cell-free DNA TF profiling is capable of detection of early-stage colorectal carcinomas. Our approach for mapping tumor-specific transcription factor binding in vivo based on blood samples makes a key part of the noncoding genome amenable to clinical analysis.

Machine learning enables detection of early-stage colorectal cancer by whole-genome sequencing of plasma cell-free DNA.

BACKGROUND: Blood-based methods using cell-free DNA (cfDNA) are under development as an alternative to existing screening tests. However, early-stage detection of cancer using tumor-derived cfDNA has proven challenging because of the small proportion of cfDNA derived from tumor tissue in early-stage disease. A machine learning approach to discover signatures in cfDNA, potentially reflective of both tumor and non-tumor contributions, may represent a promising direction for the early detection of cancer. METHODS: Whole-genome sequencing was performed on cfDNA extracted from plasma samples (N = 546 colorectal cancer and 271 non-cancer controls). Reads aligning to protein-coding gene bodies were extracted, and read counts were normalized. cfDNA tumor fraction was estimated using IchorCNA. Machine learning models were trained using k-fold cross-validation and confounder-based cross-validations to assess generalization performance. RESULTS: In a colorectal cancer cohort heavily weighted towards early-stage cancer (80% stage I/II), we achieved a mean AUC of 0.92 (95% CI 0.91-0.93) with a mean sensitivity of 85% (95% CI 83-86%) at 85% specificity. Sensitivity generally increased with tumor stage and increasing tumor fraction. Stratification by age, sequencing batch, and institution demonstrated the impact of these confounders and provided a more accurate assessment of generalization performance. CONCLUSIONS: A machine learning approach using cfDNA achieved high sensitivity and specificity in a large, predominantly early-stage, colorectal cancer cohort. The possibility of systematic technical and institution-specific biases warrants similar confounder analyses in other studies. Prospective validation of this machine learning method and evaluation of a multi-analyte approach are underway.

Reliability of the supraorbital ethmoid cell vs Keros classification in predicting the course of the anterior ethmoid artery.

BACKGROUND: We previously showed that the supraorbital ethmoid cell (SOEC) is a reliable landmark for identifying the anterior ethmoid artery (AEA). Recent data have suggested that Keros classification is also a dependable predictor. We aim to characterize the location of the AEA and its relation to the skull base in patients with and without SOEC using the Keros classification. METHODS: Retrospective radiographic evaluation of computed tomography (CT) scans of 76 patients (40 with SOEC, 36 without) was conducted. Distance of AEA from skull base and prevalence of AEA outside of the skull base were measured on each side and compared between groups using the 2-sample t test and χ2 test, respectively. Subgroup analysis was carried out based on the Keros classification. RESULTS: Mean distance of AEA from the skull base was 1.32 ± 1.5 mm in patients with SOEC and 0.47 ± 1.08 mm in those without (p < 0.001). Prevalence of AEA outside of the skull base was 53.8% in those with SOEC and 18.1% in those without (p < 0.001). Comparing patients with SOEC to those without, AEA was found below the skull base in 30% vs 0% of cases with Keros type 1 (p = 0.45), 58% vs 14.5% with Keros type 2 (p < 0.001), and 60% vs 50% with Keros type 3 (p = 0.72). CONCLUSION: The presence of SOEC is associated with a higher prevalence of the AEA coursing below the level of the skull base in all Keros types, thus placing the artery at greater risk for injury. Careful surgical planning is needed to avoid potential orbital complications.