Predicting decompression surgery by applying multimodal deep learning to patients' structured and unstructured health data.

BACKGROUND: Low back pain (LBP) is a common condition made up of a variety of anatomic and clinical subtypes. Lumbar disc herniation (LDH) and lumbar spinal stenosis (LSS) are two subtypes highly associated with LBP. Patients with LDH/LSS are often started with non-surgical treatments and if those are not effective then go on to have decompression surgery. However, recommendation of surgery is complicated as the outcome may depend on the patient's health characteristics. We developed a deep learning (DL) model to predict decompression surgery for patients with LDH/LSS. MATERIALS AND METHOD: We used datasets of 8387 and 8620 patients from a prospective study that collected data from four healthcare systems to predict early (within 2 months) and late surgery (within 12 months after a 2 month gap), respectively. We developed a DL model to use patients' demographics, diagnosis and procedure codes, drug names, and diagnostic imaging reports to predict surgery. For each prediction task, we evaluated the model's performance using classical and generalizability evaluation. For classical evaluation, we split the data into training (80%) and testing (20%). For generalizability evaluation, we split the data based on the healthcare system. We used the area under the curve (AUC) to assess performance for each evaluation. We compared results to a benchmark model (i.e. LASSO logistic regression). RESULTS: For classical performance, the DL model outperformed the benchmark model for early surgery with an AUC of 0.725 compared to 0.597. For late surgery, the DL model outperformed the benchmark model with an AUC of 0.655 compared to 0.635. For generalizability performance, the DL model outperformed the benchmark model for early surgery. For late surgery, the benchmark model outperformed the DL model. CONCLUSIONS: For early surgery, the DL model was preferred for classical and generalizability evaluation. However, for late surgery, the benchmark and DL model had comparable performance. Depending on the prediction task, the balance of performance may shift between DL and a conventional ML method. As a result, thorough assessment is needed to quantify the value of DL, a relatively computationally expensive, time-consuming and less interpretable method.

Mandibular Condylar Fractures.

Mandibular condyle fractures can result in short-term and long-term morbidity. As a weak area of the mandible, the condyle is vulnerable to injury by a direct impact or an indirect force. Current treatment recommendations aim to better match the severity of the fracture with the choice of closed or open approach. Long-term follow-up of patients provides the best opportunity to monitor the degree of functional restoration after treatment. There is a growing consensus regarding the use of standardized fracture classification methods and outcomes measures that will allow better assessment of treatment results and strengthen the quality of outcomes research.

Systematic review of submental artery island flap versus free flap in head and neck reconstruction.

PURPOSE: The aim of this systematic review is to compare the perioperative characteristics and outcomes of submental artery island flap (SAIF) to free tissue transfer (FTT) in head and neck reconstruction. MATERIALS AND METHODS: Screening and data extraction were done with Pubmed, Embase, and Web of Science databases by two independent authors to identify randomized and observational studies that compared patient outcomes for SAIF vs. FTT for reconstruction head and neck cancer ablative surgery. Data were pooled with random-effects meta-analysis to determine pooled difference in means (DM), absolute risk differences, and 95% confidence intervals (CI). Heterogeneity was assessed with the I-squared statistic. RESULTS: Initial query yielded 997 results, of which 7 studies met inclusion criteria. The pooled sample sizes for the SAIF and FTT cohorts were 155 and 198, respectively. SAIF reduced mean operative time by 193 min (95% CI -160 to -227), reduced hospital stay by 2.1 days (95% CI -0.9 to -3.4), and had a smaller flap area of 22.5cm2 (95% CI 6.5 to 38.4). SAIF had a 5% higher incidence of partial flap necrosis than FTT (95% CI, 1 to 10), but all other perioperative complications, including recurrence rate in malignant cases, were statistically comparable. CONCLUSIONS: The SAIF requires less operative time, hospital stay, and has comparable perioperative outcomes to FTT, but the area of flap harvest is significantly smaller. The findings of this study add to the growing body of evidence demonstrating the safety and reliability of SAIF in head and neck reconstruction.

Rationale and design of the Kidney Precision Medicine Project.

Chronic kidney disease (CKD) and acute kidney injury (AKI) are common, heterogeneous, and morbid diseases. Mechanistic characterization of CKD and AKI in patients may facilitate a precision-medicine approach to prevention, diagnosis, and treatment. The Kidney Precision Medicine Project aims to ethically and safely obtain kidney biopsies from participants with CKD or AKI, create a reference kidney atlas, and characterize disease subgroups to stratify patients based on molecular features of disease, clinical characteristics, and associated outcomes. An additional aim is to identify critical cells, pathways, and targets for novel therapies and preventive strategies. This project is a multicenter prospective cohort study of adults with CKD or AKI who undergo a protocol kidney biopsy for research purposes. This investigation focuses on kidney diseases that are most prevalent and therefore substantially burden the public health, including CKD attributed to diabetes or hypertension and AKI attributed to ischemic and toxic injuries. Reference kidney tissues (for example, living-donor kidney biopsies) will also be evaluated. Traditional and digital pathology will be combined with transcriptomic, proteomic, and metabolomic analysis of the kidney tissue as well as deep clinical phenotyping for supervised and unsupervised subgroup analysis and systems biology analysis. Participants will be followed prospectively for 10 years to ascertain clinical outcomes. Cell types, locations, and functions will be characterized in health and disease in an open, searchable, online kidney tissue atlas. All data from the Kidney Precision Medicine Project will be made readily available for broad use by scientists, clinicians, and patients.

Modelling kidney disease using ontology: insights from the Kidney Precision Medicine Project.

An important need exists to better understand and stratify kidney disease according to its underlying pathophysiology in order to develop more precise and effective therapeutic agents. National collaborative efforts such as the Kidney Precision Medicine Project are working towards this goal through the collection and integration of large, disparate clinical, biological and imaging data from patients with kidney disease. Ontologies are powerful tools that facilitate these efforts by enabling researchers to organize and make sense of different data elements and the relationships between them. Ontologies are critical to support the types of big data analysis necessary for kidney precision medicine, where heterogeneous clinical, imaging and biopsy data from diverse sources must be combined to define a patient's phenotype. The development of two new ontologies - the Kidney Tissue Atlas Ontology and the Ontology of Precision Medicine and Investigation - will support the creation of the Kidney Tissue Atlas, which aims to provide a comprehensive molecular, cellular and anatomical map of the kidney. These ontologies will improve the annotation of kidney-relevant data, and eventually lead to new definitions of kidney disease in support of precision medicine.

Comparing Perioperative Outcomes of Uterine Artery Embolization and Hysterectomy in Insurer and Demographically Diverse Populations: A Retrospective, Multi-Center Database Study.

INTRODUCTION: Past studies comparing perioperative outcomes of hysterectomy (HYST) and uterine artery embolization (UAE) do not control for demographically and insurer diverse populations. This study sought to identify the 30­day readmission, 15­day complication, and minimum 1­year surveillance reintervention rates of diverse, propensity matched patients undergoing UAE or HYST for uterine leiomyoma. METHODS: Adults from the New York's Statewide Planning and Research Cooperative System (SPARCS) database 2009­2013 who underwent HYST or UAE for uterine leiomyoma were retrospectively reviewed and 1:1 propensity matched. Univariate analysis compared demographics, complications, readmissions, and reintervention rates. Binary logistic and linear regression models identified independent predictors of outcomes. RESULTS: A total of 682 patients were identified, where the number (n) of patients in each cohort was n=341, HYST, and n=341, UAE. Significance levels are shown with p values. No significant differences were identified between HYST and UAE demographics, complication (2.60% HYST vs 2.90% UAE, p=0.816) or readmission rates (3.20% HYST vs 3.80% UAE, p=0.678); 0.30% of UAE patients had a reintervention UAE and 2.90% of UAE patients had reintervention hysterectomy. HYST patients had a significantly longer average length of stay (2.59 days HYST vs 1.63 days UAE, p<0.001). The Deyo-Charlson (Deyo) comorbidity score positively predicted any complication with odds ratio=34.262, 95% confidence interval [4.938, 237.725], and p<0.001, but did not predict readmissions. CONCLUSION: HYST patients had significantly longer hospital stays. UAE and HYST had comparable readmission and complication rates. The Deyo comorbidity score was a significant predictor of any complication. This study supports the safety and efficacy of UAE when compared to HYST in demographic and insurer diverse populations.

Assessing computational predictions of the phenotypic effect of cystathionine-beta-synthase variants.

Accurate prediction of the impact of genomic variation on phenotype is a major goal of computational biology and an important contributor to personalized medicine. Computational predictions can lead to a better understanding of the mechanisms underlying genetic diseases, including cancer, but their adoption requires thorough and unbiased assessment. Cystathionine-beta-synthase (CBS) is an enzyme that catalyzes the first step of the transsulfuration pathway, from homocysteine to cystathionine, and in which variations are associated with human hyperhomocysteinemia and homocystinuria. We have created a computational challenge under the CAGI framework to evaluate how well different methods can predict the phenotypic effect(s) of CBS single amino acid substitutions using a blinded experimental data set. CAGI participants were asked to predict yeast growth based on the identity of the mutations. The performance of the methods was evaluated using several metrics. The CBS challenge highlighted the difficulty of predicting the phenotype of an ex vivo system in a model organism when classification models were trained on human disease data. We also discuss the variations in difficulty of prediction for known benign and deleterious variants, as well as identify methodological and experimental constraints with lessons to be learned for future challenges.

Assessment of blind predictions of the clinical significance of BRCA1 and BRCA2 variants.

Testing for variation in BRCA1 and BRCA2 (commonly referred to as BRCA1/2), has emerged as a standard clinical practice and is helping countless women better understand and manage their heritable risk of breast and ovarian cancer. Yet the increased rate of BRCA1/2 testing has led to an increasing number of Variants of Uncertain Significance (VUS), and the rate of VUS discovery currently outpaces the rate of clinical variant interpretation. Computational prediction is a key component of the variant interpretation pipeline. In the CAGI5 ENIGMA Challenge, six prediction teams submitted predictions on 326 newly-interpreted variants from the ENIGMA Consortium. By evaluating these predictions against the new interpretations, we have gained a number of insights on the state of the art of variant prediction and specific steps to further advance this state of the art.

Pathogenicity and functional impact of non-frameshifting insertion/deletion variation in the human genome.

Differentiation between phenotypically neutral and disease-causing genetic variation remains an open and relevant problem. Among different types of variation, non-frameshifting insertions and deletions (indels) represent an understudied group with widespread phenotypic consequences. To address this challenge, we present a machine learning method, MutPred-Indel, that predicts pathogenicity and identifies types of functional residues impacted by non-frameshifting insertion/deletion variation. The model shows good predictive performance as well as the ability to identify impacted structural and functional residues including secondary structure, intrinsic disorder, metal and macromolecular binding, post-translational modifications, allosteric sites, and catalytic residues. We identify structural and functional mechanisms impacted preferentially by germline variation from the Human Gene Mutation Database, recurrent somatic variation from COSMIC in the context of different cancers, as well as de novo variants from families with autism spectrum disorder. Further, the distributions of pathogenicity prediction scores generated by MutPred-Indel are shown to differentiate highly recurrent from non-recurrent somatic variation. Collectively, we present a framework to facilitate the interrogation of both pathogenicity and the functional effects of non-frameshifting insertion/deletion variants. The MutPred-Indel webserver is available at http://mutpred.mutdb.org/.

Assessing the performance of in silico methods for predicting the pathogenicity of variants in the gene CHEK2, among Hispanic females with breast cancer.

The availability of disease-specific genomic data is critical for developing new computational methods that predict the pathogenicity of human variants and advance the field of precision medicine. However, the lack of gold standards to properly train and benchmark such methods is one of the greatest challenges in the field. In response to this challenge, the scientific community is invited to participate in the Critical Assessment for Genome Interpretation (CAGI), where unpublished disease variants are available for classification by in silico methods. As part of the CAGI-5 challenge, we evaluated the performance of 18 submissions and three additional methods in predicting the pathogenicity of single nucleotide variants (SNVs) in checkpoint kinase 2 (CHEK2) for cases of breast cancer in Hispanic females. As part of the assessment, the efficacy of the analysis method and the setup of the challenge were also considered. The results indicated that though the challenge could benefit from additional participant data, the combined generalized linear model analysis and odds of pathogenicity analysis provided a framework to evaluate the methods submitted for SNV pathogenicity identification and for comparison to other available methods. The outcome of this challenge and the approaches used can help guide further advancements in identifying SNV-disease relationships.

Working toward precision medicine: Predicting phenotypes from exomes in the Critical Assessment of Genome Interpretation (CAGI) challenges.

Precision medicine aims to predict a patient's disease risk and best therapeutic options by using that individual's genetic sequencing data. The Critical Assessment of Genome Interpretation (CAGI) is a community experiment consisting of genotype-phenotype prediction challenges; participants build models, undergo assessment, and share key findings. For CAGI 4, three challenges involved using exome-sequencing data: Crohn's disease, bipolar disorder, and warfarin dosing. Previous CAGI challenges included prior versions of the Crohn's disease challenge. Here, we discuss the range of techniques used for phenotype prediction as well as the methods used for assessing predictive models. Additionally, we outline some of the difficulties associated with making predictions and evaluating them. The lessons learned from the exome challenges can be applied to both research and clinical efforts to improve phenotype prediction from genotype. In addition, these challenges serve as a vehicle for sharing clinical and research exome data in a secure manner with scientists who have a broad range of expertise, contributing to a collaborative effort to advance our understanding of genotype-phenotype relationships.

CADD score has limited clinical validity for the identification of pathogenic variants in noncoding regions in a hereditary cancer panel.

PURPOSE: Several in silico tools have been shown to have reasonable research sensitivity and specificity for classifying sequence variants in coding regions. The recently developed combined annotation-dependent depletion (CADD) method generates predictive scores for single-nucleotide variants (SNVs) in all areas of the genome, including noncoding regions. We sought for non-coding variants to determine the clinical validity of common CADD scores. METHODS: We evaluated 12,391 unique SNVs in 624 patient samples submitted for germ-line mutation testing in a cancer-related gene panel. Stratifying by genomic region, we compared the distributions of CADD scores of rare SNVs, SNVs common in our patient population, and the null distribution of all possible SNVs. RESULTS: The median CADD scores of intronic and nonsynonymous variants were significantly different between rare and common SNVs (P < 0.0001). Despite these different distributions, no individual variants could be identified as plausibly causative among the rare intronic variants with the highest scores. The receiver-operating characteristics (ROC) area under the curve (AUC) for noncoding variants is modest, and the positive predictive value of CADD for intronic variants in panel testing was found to be 0.088. CONCLUSION: Focused in silico scoring systems with much higher predictive value will be necessary for clinical genomic applications.Genet Med 18 12, 1269-1275.

Genomic Analysis Reveals Disruption of Striatal Neuronal Development and Therapeutic Targets in Human Huntington's Disease Neural Stem Cells.

We utilized induced pluripotent stem cells (iPSCs) derived from Huntington's disease (HD) patients as a human model of HD and determined that the disease phenotypes only manifest in the differentiated neural stem cell (NSC) stage, not in iPSCs. To understand the molecular basis for the CAG repeat expansion-dependent disease phenotypes in NSCs, we performed transcriptomic analysis of HD iPSCs and HD NSCs compared to isogenic controls. Differential gene expression and pathway analysis pointed to transforming growth factor ß (TGF-ß) and netrin-1 as the top dysregulated pathways. Using data-driven gene coexpression network analysis, we identified seven distinct coexpression modules and focused on two that were correlated with changes in gene expression due to the CAG expansion. Our HD NSC model revealed the dysregulation of genes involved in neuronal development and the formation of the dorsal striatum. The striatal and neuronal networks disrupted could be modulated to correct HD phenotypes and provide therapeutic targets.

A probabilistic model to predict clinical phenotypic traits from genome sequencing.

Genetic screening is becoming possible on an unprecedented scale. However, its utility remains controversial. Although most variant genotypes cannot be easily interpreted, many individuals nevertheless attempt to interpret their genetic information. Initiatives such as the Personal Genome Project (PGP) and Illumina's Understand Your Genome are sequencing thousands of adults, collecting phenotypic information and developing computational pipelines to identify the most important variant genotypes harbored by each individual. These pipelines consider database and allele frequency annotations and bioinformatics classifications. We propose that the next step will be to integrate these different sources of information to estimate the probability that a given individual has specific phenotypes of clinical interest. To this end, we have designed a Bayesian probabilistic model to predict the probability of dichotomous phenotypes. When applied to a cohort from PGP, predictions of Gilbert syndrome, Graves' disease, non-Hodgkin lymphoma, and various blood groups were accurate, as individuals manifesting the phenotype in question exhibited the highest, or among the highest, predicted probabilities. Thirty-eight PGP phenotypes (26%) were predicted with area-under-the-ROC curve (AUC)>0.7, and 23 (15.8%) of these were statistically significant, based on permutation tests. Moreover, in a Critical Assessment of Genome Interpretation (CAGI) blinded prediction experiment, the models were used to match 77 PGP genomes to phenotypic profiles, generating the most accurate prediction of 16 submissions, according to an independent assessor. Although the models are currently insufficiently accurate for diagnostic utility, we expect their performance to improve with growth of publicly available genomics data and model refinement by domain experts.

MutPred Splice: machine learning-based prediction of exonic variants that disrupt splicing.

We have developed a novel machine-learning approach, MutPred Splice, for the identification of coding region substitutions that disrupt pre-mRNA splicing. Applying MutPred Splice to human disease-causing exonic mutations suggests that 16% of mutations causing inherited disease and 10 to 14% of somatic mutations in cancer may disrupt pre-mRNA splicing. For inherited disease, the main mechanism responsible for the splicing defect is splice site loss, whereas for cancer the predominant mechanism of splicing disruption is predicted to be exon skipping via loss of exonic splicing enhancers or gain of exonic splicing silencer elements. MutPred Splice is available at http://mutdb.org/mutpredsplice.

Intrinsic Size Parameters for Palmitoylated and Carboxyamidomethylated Peptides.

Cross sections for 61 palmitoylated peptides and 73 cysteine-unmodified peptides are determined and used together with a previously obtained tryptic peptide library to derive a set of intrinsic size parameters (ISPs) for the palmitoyl (Pal) group (1.26 ± 0.04), carboxyamidomethyl (Am) group (0.92 ± 0.04), and the 20 amino acid residues to assess the influence of Pal- and Am-modification on cysteine and other amino acid residues. These values highlight the influence of the intrinsic hydrophobic and hydrophilic nature of these modifications on the overall cross sections. As a part of this analysis, we find that ISPs derived from a database of a modifier on one amino acid residue (CysPal) can be applied on the same modification group on different amino acid residues (SerPal and TyrPal). Using these ISP values, we are able to calculate peptide cross sections to within ± 2% of experimental values for 83% of Pal-modified peptide ions and 63% of Am-modified peptide ions. We propose that modification groups should be treated as individual contribution factors, instead of treating the combination of the particular group and the amino acid residue they are on as a whole when considering their effects on the peptide ion mobility features.

Calibration of multiple in silico tools for predicting pathogenicity of mismatch repair gene missense substitutions.

Classification of rare missense substitutions observed during genetic testing for patient management is a considerable problem in clinical genetics. The Bayesian integrated evaluation of unclassified variants is a solution originally developed for BRCA1/2. Here, we take a step toward an analogous system for the mismatch repair (MMR) genes (MLH1, MSH2, MSH6, and PMS2) that confer colon cancer susceptibility in Lynch syndrome by calibrating in silico tools to estimate prior probabilities of pathogenicity for MMR gene missense substitutions. A qualitative five-class classification system was developed and applied to 143 MMR missense variants. This identified 74 missense substitutions suitable for calibration. These substitutions were scored using six different in silico tools (Align-Grantham Variation Grantham Deviation, multivariate analysis of protein polymorphisms [MAPP], MutPred, PolyPhen-2.1, Sorting Intolerant From Tolerant, and Xvar), using curated MMR multiple sequence alignments where possible. The output from each tool was calibrated by regression against the classifications of the 74 missense substitutions; these calibrated outputs are interpretable as prior probabilities of pathogenicity. MAPP was the most accurate tool and MAPP + PolyPhen-2.1 provided the best-combined model (R(2)  = 0.62 and area under receiver operating characteristic = 0.93). The MAPP + PolyPhen-2.1 output is sufficiently predictive to feed as a continuous variable into the quantitative Bayesian integrated evaluation for clinical classification of MMR gene missense substitutions.

An excess of deleterious variants in VEGF-A pathway genes in Down-syndrome-associated atrioventricular septal defects.

About half of people with trisomy 21 have a congenital heart defect (CHD), whereas the remainder have a structurally normal heart, demonstrating that trisomy 21 is a significant risk factor but is not causal for abnormal heart development. Atrioventricular septal defects (AVSD) are the most commonly occurring heart defects in Down syndrome (DS), and ∼65% of all AVSD is associated with DS. We used a candidate-gene approach among individuals with DS and complete AVSD (cases = 141) and DS with no CHD (controls = 141) to determine whether rare genetic variants in genes involved in atrioventricular valvuloseptal morphogenesis contribute to AVSD in this sensitized population. We found a significant excess (p < 0.0001) of variants predicted to be deleterious in cases compared to controls. At the most stringent level of filtering, we found potentially damaging variants in nearly 20% of cases but fewer than 3% of controls. The variants with the highest probability of being damaging in cases only were found in six genes: COL6A1, COL6A2, CRELD1, FBLN2, FRZB, and GATA5. Several of the case-specific variants were recurrent in unrelated individuals, occurring in 10% of cases studied. No variants with an equal probability of being damaging were found in controls, demonstrating a highly specific association with AVSD. Of note, all of these genes are in the VEGF-A pathway, even though the candidate genes analyzed in this study represented numerous biochemical and developmental pathways, suggesting that rare variants in the VEGF-A pathway might contribute to the genetic underpinnings of AVSD in humans.

Characterization of ligand type of estrogen receptor by MD simulation and mm-PBSA free energy analysis.

Estrogen receptor is a transcription regulator and can bind structurally distinct ligands with full agonistic, SERMs, or full antagonistic properties. Crystal structures of the ER ligand binding domain (LBD)-complexed with full agonists or SERMs show that these ligands induce two different orientations of Helix12 in LBD and generate two different conformations, agonist conformation (A conformation) and AF2 antagonist conformation (B conformation). To understand how ER ligands interact with LBD structurally and energetically, we docked 3 full agonists, 9 SERMs and 2 full antagonists in both the A and B conformation of ERα LBD and performed a 4-step molecular dynamics (MD) simulation on all 28 complexes followed by mm-PBSA binding free energy calculation. We found that all full agonists prefer the A conformation while all SERMs prefer the B conformation. Analysis of the mm-PBSA energies revealed that calculated total binding free energies (delta PBTOT) and the difference of VDW between complex and the sum of receptor of ligands and ligand (delta VDW) have the order of full agonists>SERMs>full antagonists. However, the PB surface term has the order of full antagonists>SERMs>full agonists. We also found that the sum of the RMSD of mainchain atoms of Helix12 and all atoms of ligands in the A conformation is significantly lower for full agonists than that of the other ligands. Together, we conclude that the three types of ER ligands interact with the A and B conformations of ERα LBD differently and same type of ligands interact similarly. These findings will be useful in understanding the mechanism of ER antagonism and can be used in ligand type prediction.

Identifying viral integration sites using SeqMap 2.0.

UNLABELLED: Retroviral integration has been implicated in several biomedical applications, including identification of cancer-associated genes and malignant transformation in gene therapy clinical trials. We introduce an efficient and scalable method for fast identification of viral vector integration sites from long read high-throughput sequencing. Individual sequence reads are masked to remove non-genomic sequence, aligned to the host genome and assembled into contiguous fragments used to pinpoint the position of integration. AVAILABILITY AND IMPLEMENTATION: The method is implemented in a publicly accessible web server platform, SeqMap 2.0, containing analysis tools and both private and shared lab workspaces that facilitate collaboration among researchers. Available at http://seqmap.compbio.iupui.edu/.