Machine Learning Methods for Identifying Atrial Fibrillation Cases and Their Predictors in Patients With Hypertrophic Cardiomyopathy: The HCM-AF-Risk Model.

BACKGROUND: Hypertrophic cardiomyopathy (HCM) patients have a high incidence of atrial fibrillation (AF) and increased stroke risk, even with low CHA2DS2-VASc (congestive heart failure, hypertension, age diabetes, previous stroke/transient ischemic attack) scores. Hence, there is a need to understand the pathophysiology of AF/stroke in HCM. In this retrospective study, we develop and apply a data-driven, machine learning-based method to identify AF cases, and clinical/imaging features associated with AF, using electronic health record data. METHODS: HCM patients with documented paroxysmal/persistent/permanent AF (n = 191) were considered AF cases, and the remaining patients in sinus rhythm (n = 640) were tagged as No-AF. We evaluated 93 clinical variables; the most informative variables useful for distinguishing AF from No-AF cases were selected based on the 2-sample t test and the information gain criterion. RESULTS: We identified 18 highly informative variables that are positively (n = 11) and negatively (n = 7) correlated with AF in HCM. Next, patient records were represented via these 18 variables. Data imbalance resulting from the relatively low number of AF cases was addressed via a combination of oversampling and undersampling strategies. We trained and tested multiple classifiers under this sampling approach, showing effective classification. Specifically, an ensemble of logistic regression and naïve Bayes classifiers, trained based on the 18 variables and corrected for data imbalance, proved most effective for separating AF from No-AF cases (sensitivity = 0.74, specificity = 0.70, C-index = 0.80). CONCLUSIONS: Our model (HCM-AF-Risk Model) is the first machine learning-based method for identification of AF cases in HCM. This model demonstrates good performance, addresses data imbalance, and suggests that AF is associated with a more severe cardiac HCM phenotype.

INTRODUCTION: Les patients atteints d'une cardiomyopathie hypertrophique (CMH) présentent une forte incidence de fibrillation auriculaire (FA) et un risque accru d'accident vasculaire cérébral (AVC), malgré des scores CHA2DS2-VASc (congestive heart failure, hypertension, age diabetes, previous stroke/transient ischemic attack, c'est-à-dire : insuffisance cardiaque congestive, hypertension, âge, diabète, AVC ou accident ischémique transitoire antérieur) faibles. Par conséquent, il est nécessaire de comprendre la physiopathologie de la FA et de l'AVC en présence d'une CMH. Dans la présente étude rétrospective, nous avons élaboré et appliqué une méthode d'apprentissage automatique dirigée sur les données pour déterminer les cas de FA, et les caractéristiques cliniques/d'imagerie associées à la FA, à l'aide des données des dossiers de santé électroniques. MÉTHODES: Nous avons considéré les patients atteints d'une CMH qui ont une FA paroxystique/persistante/permanente documentée (n = 191) comme des cas de FA, et avons étiqueté les autres patients en rythme sinusal (n = 640) comme des cas sans FA. Nous avons évalué 93 variables cliniques; nous avons sélectionné les variables les plus informatives qui sont utiles pour distinguer les cas de FA des cas sans FA en fonction du test t pour deux échantillons et du critère de gain d'information. RÉSULTATS: Nous avons relevé 18 variables hautement informatives qui ont une corrélation positive (n = 11) et une corrélation négative (n = 7) avec la FA en présence d'une CMH. Ensuite, nous avons représenté les dossiers des patients au moyen de ces 18 variables. Nous avons remédié au déséquilibre des données, qui résulte du nombre relativement faible de cas de FA, grâce à une combinaison de stratégies de suréchantillonnage et de sous-échantillonnage. Nous avons formé et testé de nombreux classificateurs selon cette approche d'échantillonnage, qui montre une classification efficace. Particulièrement, un ensemble de régression logistique et de classificateurs bayésiens naïfs formés en fonction des 18 variables et corrigés en fonction du déséquilibre des données s'est révélé le plus efficace pour séparer les cas de FA des cas sans FA (sensibilité = 0,74, spécificité = 0,70, indice C = 0,80). CONCLUSIONS: Notre modèle (modèle de risque de CMH-FA) est la première méthode d'apprentissage automatique qui sert à déterminer les cas de FA en présence de CMH. Ce modèle permet de démontrer une bonne performance, de remédier au déséquilibre des données, et de croire que la FA est associée à un phénotype grave de CMH.

Hydrogenation and electrocatalytic reduction of carbon dioxide to formate with a single Co catalyst.

A cobalt(i) complex is shown to be capable of both electrocatalytic reduction and hydrogenation of CO2 to formate. Several proposed intermediates are characterized and thus form the basis for a proposed mechanism that allows for the dual reactivity: reduction of CO2via H2 addition, and H+/e- equivalents. The work makes use of a novel tris(phosphino) ligand. When a pendent amine is attached to the ligand, no change in catalytic reactivity is observed.

Toward Combined Carbon Capture and Recycling: Addition of an Amine Alters Product Selectivity from CO to Formic Acid in Manganese Catalyzed Reduction of CO2.

Owing to the energetic cost associated with CO2 release in carbon capture (CC), the combination of carbon capture and recycling (CCR) is an emerging area of research. In this approach, "captured CO2," typically generated by addition of amines, serves as a substrate for subsequent reduction. Herein, we report that the reduction of CO2 in the presence of morpholine (generating mixtures of the corresponding carbamate and carbamic acid) with a well-established Mn electrocatalyst changes the product selectivity from CO to H2 and formate. The change in selectivity is attributed to in situ generation of the morpholinium carbamic acid, which is sufficiently acidic to protonate the reduced Mn species and generate an intermediate Mn hydride. Thermodynamic studies indicate that the hydride is not sufficiently hydritic to reduce CO2 to formate, unless the apparent hydricity, which encompasses formate binding to the Mn, is considered. Increasing steric bulk around the Mn shuts down rapid homolytic H2 evolution rendering the intermediate Mn hydride more stable; subsequent CO2 insertion appears to be faster than heterolytic H2 production. A comprehensive mechanistic scheme is proposed that illustrates how thermodynamic analysis can provide further insight. Relevant to a range of hydrogenations and reductions is the modulation of the hydricity with substrate binding that makes the reaction favorable. Significantly, this work illustrates a new role for amines in CO2 reduction: changing the product selectivity; this is pertinent more broadly to advancing CCR.

The role of leached Zr in the photocatalytic reduction of CO2 to formate by derivatives of UiO-66 metal organic frameworks.

This work reports the photocatalytic reduction of CO2 to formate that is mediated by UiO-66 Zr MOF derivatives. Amino-substituted UiO-66 is a known photocatalyst for this transformation, and herein we identify that this catalysis is due to leached Zr, not the MOF itself. No correlation between catalytic activity and crystal size is observed for UiO-66-NH2. Recycling studies along with SEM images of the crystals prior to and after catalysis support our conclusion that the catalytic performance can be described by the amount of leached Zr. Moreover, when the effect of the linker on the catalytic reaction is probed, all MOFs that facilitate the reduction of CO2 are found to leach Zr into solution. Correlation of the MOF (or linker) band gap energies to formate production indicates that this is an important parameter to the leached species. Combined with a linker exchange study, this indicates that the leached Zr still coordinates the linker. These results indicate that the UiO-66 Zr MOFs are not stable under typical photochemical conditions, and emphasize the importance of considering the role that leached metals play in catalysis.

Identifying Ventricular Arrhythmias and Their Predictors by Applying Machine Learning Methods to Electronic Health Records in Patients With Hypertrophic Cardiomyopathy (HCM-VAr-Risk Model).

Clinical risk stratification for sudden cardiac death (SCD) in hypertrophic cardiomyopathy (HC) employs rules derived from American College of Cardiology Foundation/American Heart Association (ACCF/AHA) guidelines or the HCM Risk-SCD model (C-index ∼0.69), which utilize a few clinical variables. We assessed whether data-driven machine learning methods that consider a wider range of variables can effectively identify HC patients with ventricular arrhythmias (VAr) that lead to SCD. We scanned the electronic health records of 711 HC patients for sustained ventricular tachycardia or ventricular fibrillation. Patients with ventricular tachycardia or ventricular fibrillation (n = 61) were tagged as VAr cases and the remaining (n = 650) as non-VAr. The 2-sample ttest and information gain criterion were used to identify the most informative clinical variables that distinguish VAr from non-VAr; patient records were reduced to include only these variables. Data imbalance stemming from low number of VAr cases was addressed by applying a combination of over- and undersampling strategies. We trained and tested multiple classifiers under this sampling approach, showing effective classification. We evaluated 93 clinical variables, of which 22 proved predictive of VAr. The ensemble of logistic regression and naïve Bayes classifiers, trained based on these 22 variables and corrected for data imbalance, was most effective in separating VAr from non-VAr cases (sensitivity = 0.73, specificity = 0.76, C-index = 0.83). Our method (HCM-VAr-Risk Model) identified 12 new predictors of VAr, in addition to 10 established SCD predictors. In conclusion, this is the first application of machine learning for identifying HC patients with VAr, using clinical attributes. Our model demonstrates good performance (C-index) compared with currently employed SCD prediction algorithms, while addressing imbalance inherent in clinical data.

Chronic Kidney Disease stratification using office visit records: Handling data imbalance via hierarchical meta-classification.

BACKGROUND: Chronic Kidney Disease (CKD) is one of several conditions that affect a growing percentage of the US population; the disease is accompanied by multiple co-morbidities, and is hard to diagnose in-and-of itself. In its advanced forms it carries severe outcomes and can lead to death. It is thus important to detect the disease as early as possible, which can help devise effective intervention and treatment plan. Here we investigate ways to utilize information available in electronic health records (EHRs) from regular office visits of more than 13,000 patients, in order to distinguish among several stages of the disease. While clinical data stored in EHRs provide valuable information for risk-stratification, one of the major challenges in using them arises from data imbalance. That is, records associated with a more severe condition are typically under-represented compared to those associated with a milder manifestation of the disease. To address imbalance, we propose and develop a sampling-based ensemble approach, hierarchical meta-classification, aiming to stratify CKD patients into severity stages, using simple quantitative non-text features gathered from standard office visit records. METHODS: The proposed hierarchical meta-classification method frames the multiclass classification task as a hierarchy of two subtasks. The first is binary classification, separating records associated with the majority class from those associated with all minority classes combined, using meta-classification. The second subtask separates the records assigned to the combined minority classes into the individual constituent classes. RESULTS: The proposed method identifies a significant proportion of patients suffering from the more advanced stages of the condition, while also correctly identifying most of the less severe cases, maintaining high sensitivity, specificity and F-measure (≥ 93%). Our results show that the high level of performance attained by our method is preserved even when the size of the training set is significantly reduced, demonstrating the stability and generalizability of our approach. CONCLUSION: We present a new approach to perform classification while addressing data imbalance, which is inherent in the biomedical domain. Our model effectively identifies severity stages of CKD patients, using information readily available in office visit records within the realistic context of high data imbalance.

Co-occurrence of medical conditions: Exposing patterns through probabilistic topic modeling of snomed codes.

Patients associated with multiple co-occurring health conditions often face aggravated complications and less favorable outcomes. Co-occurring conditions are especially prevalent among individuals suffering from kidney disease, an increasingly widespread condition affecting 13% of the general population in the US. This study aims to identify and characterize patterns of co-occurring medical conditions in patients employing a probabilistic framework. Specifically, we apply topic modeling in a non-traditional way to find associations across SNOMED-CT codes assigned and recorded in the EHRs of >13,000 patients diagnosed with kidney disease. Unlike most prior work on topic modeling, we apply the method to codes rather than to natural language. Moreover, we quantitatively evaluate the topics, assessing their tightness and distinctiveness, and also assess the medical validity of our results. Our experiments show that each topic is succinctly characterized by a few highly probable and unique disease codes, indicating that the topics are tight. Furthermore, inter-topic distance between each pair of topics is typically high, illustrating distinctiveness. Last, most coded conditions grouped together within a topic, are indeed reported to co-occur in the medical literature. Notably, our results uncover a few indirect associations among conditions that have hitherto not been reported as correlated in the medical literature.

Modulation of curli assembly and pellicle biofilm formation by chemical and protein chaperones.

Enteric bacteria assemble functional amyloid fibers, curli, on their surfaces that share structural and biochemical properties with disease-associated amyloids. Here, we test rationally designed 2-pyridone compounds for their ability to alter amyloid formation of the major curli subunit CsgA. We identified several compounds that discourage CsgA amyloid formation and several compounds that accelerate CsgA amyloid formation. The ability of inhibitor compounds to stop growing CsgA fibers was compared to the same property of the CsgA chaperone, CsgE. CsgE blocked CsgA amyloid assembly and arrested polymerization when added to actively polymerizing fibers. Additionally, CsgE and the 2-pyridone inhibitors prevented biofilm formation by Escherichia coli at the air-liquid interface of a static culture. We demonstrate that curli amyloid assembly and curli-dependent biofilm formation can be modulated not only by protein chaperones, but also by "chemical chaperones."

Rational design of potent domain antibody inhibitors of amyloid fibril assembly.

Antibodies hold significant potential for inhibiting toxic protein aggregation associated with conformational disorders such as Alzheimer's and Huntington's diseases. However, near-stoichiometric antibody concentrations are typically required to completely inhibit protein aggregation. We posited that the molecular interactions mediating amyloid fibril formation could be harnessed to generate antibodies with potent antiaggregation. Here we report that grafting small amyloidogenic peptides (6-10 residues) into the complementarity-determining regions of a single-domain (V(H)) antibody yields potent domain antibody inhibitors of amyloid formation. Grafted AMyloid-Motif AntiBODIES (gammabodies) presenting hydrophobic peptides from Aß (Alzheimer's disease), α-Synuclein (Parkinson's disease), and islet amyloid polypeptide (type 2 diabetes) inhibit fibril assembly of each corresponding polypeptide at low substoichiometric concentrations (1:10 gammabody:monomer molar ratio). In contrast, sequence- and conformation-specific antibodies that were obtained via immunization are unable to prevent fibrillization at the same substoichiometric concentrations. Gammabodies prevent amyloid formation by converting monomers and/or fibrillar intermediates into small complexes that are unstructured and benign. We expect that our antibody design approach--which eliminates the need for immunization or screening to identify sequence-specific domain antibody inhibitors--can be readily extended to generate potent aggregation inhibitors of other amyloidogenic polypeptides linked to human disease.

Aggregation-resistant domain antibodies engineered with charged mutations near the edges of the complementarity-determining regions.

Antibodies commonly contain hydrophobic residues within their complementarity-determining regions (CDRs) that mediate binding to target antigens. Unfortunately, hydrophobic CDRs can also promote antibody aggregation, which is especially concerning for therapeutic antibodies due to the immunogenicity of antibody aggregates. Here we investigate how the sequences of CDRs within single-domain (V(H)) antibodies specific for the Alzheimer's amyloid ß peptide can be engineered to resist aggregation without reducing binding affinity. We find that domain antibodies containing clusters of hydrophobic residues within their third CDR (CDR3) are prone to aggregate within days at 25°C and minutes above 70°C. However, inserting two or more negatively charged residues at each edge of CDR3 potently suppresses antibody aggregation without altering binding affinity. We also find that inserting charged mutations at one edge of CDR3 (N- or C-terminal) prevents aggregation, but only if such mutations are located at the edge closest to most hydrophobic portion of CDR3. In contrast, charged mutations outside of CDR3 fail to suppress aggregation. Our findings demonstrate that the sequence of CDR loops can be engineered in a systematic manner to improve antibody solubility without altering binding affinity or specificity.

Polyphenolic disaccharides endow proteins with unusual resistance to aggregation.

Protein aggregation is a common problem during the purification and formulation of therapeutic proteins. Here we report that polyphenolic disaccharides are unusually effective at preventing protein aggregation. We find that two polyphenolic glycosides-naringin and rutin-endow diverse proteins with the ability to unfold without aggregating when heated, as well as the ability to refold without aggregating when cooled at low glycoside concentrations (<5 mM). This extreme solubilizing activity is a synergistic combination of the glycone and aglycone moieties, as combinations of polyphenols and sugars fail to suppress aggregation. Moreover, the activity of polyphenolic disaccharides is remarkably specific since their monosaccharide counterparts (as well as other common excipients such as arginine, trehalose, and cyclodextrin) fail to prevent aggregation at similar concentrations (<25 mM). We expect that polyphenolic disaccharides will be valuable additives for enhancing the solubility of proteins in applications plagued by protein aggregation.

Structure-based design of conformation- and sequence-specific antibodies against amyloid ß.

Conformation-specific antibodies that recognize aggregated proteins associated with several conformational disorders (e.g., Parkinson and prion diseases) are invaluable for diagnostic and therapeutic applications. However, no systematic strategy exists for generating conformation-specific antibodies that target linear sequence epitopes within misfolded proteins. Here we report a strategy for designing conformation- and sequence-specific antibodies against misfolded proteins that is inspired by the molecular interactions governing protein aggregation. We find that grafting small amyloidogenic peptides (6-10 residues) from the Aß42 peptide associated with Alzheimer's disease into the complementarity determining regions of a domain (V(H)) antibody generates antibody variants that recognize Aß soluble oligomers and amyloid fibrils with nanomolar affinity. We refer to these antibodies as gammabodies for grafted amyloid-motif antibodies. Gammabodies displaying the central amyloidogenic Aß motif (18VFFA21) are reactive with Aß fibrils, whereas those displaying the amyloidogenic C terminus (34LMVGGVVIA42) are reactive with Aß fibrils and oligomers (and weakly reactive with Aß monomers). Importantly, we find that the grafted motifs target the corresponding peptide segments within misfolded Aß conformers. Aß gammabodies fail to cross-react with other amyloidogenic proteins and scrambling their grafted sequences eliminates antibody reactivity. Finally, gammabodies that recognize Aß soluble oligomers and fibrils also neutralize the toxicity of each Aß conformer. We expect that our antibody design strategy is not limited to Aß and can be used to readily generate gammabodies against other toxic misfolded proteins.

Site-specific structural analysis of a yeast prion strain with species-specific seeding activity.

Prion proteins misfold and aggregate into multiple infectious strain variants that possess unique abilities to overcome prion species barriers, yet the structural basis for the species-specific infectivities of prion strains is poorly understood. Therefore, we have investigated the site-specific structural properties of a promiscuous chimeric form of the yeast prion Sup35 from Saccharomyces cerevisiae and Candida albicans. The Sup35 chimera forms two strain variants, each of which selectively infect one species but not the other. Importantly, the N-terminal and middle domains of the Sup35 chimera (collectively referred to as Sup35NM) contain two prion recognition elements (one from each species) that regulate the nucleation of each strain. Mutations in either prion recognition element significantly bias nucleation of one strain conformation relative to the other. Herein, we have investigated the folding of each prion recognition element for the serine-to-arginine mutant at residue 17 of Sup35NM chimera known to promote nucleation of C. albicans strain conformation. Using cysteine-specific labeling analysis, we find that residues in the C. albicans prion recognition element are solvent-shielded, while those outside the recognition sequence (including most of those in the S. cerevisiae recognition element) are solvent-exposed. Moreover, we find that proline mutations in the C. albicans recognition sequence disrupt the prion templating activity of this strain conformation. Our structural findings reveal that differential folding of complementary and non-complementary prion recognition elements within the prion amyloid core of the Sup35NM chimera is the structural basis for its species-specific templating activity.

Mutational analysis of domain antibodies reveals aggregation hotspots within and near the complementarity determining regions.

High-affinity antibodies are critical for numerous diagnostic and therapeutic applications, yet their utility is limited by their variable propensity to aggregate either at low concentrations for antibody fragments or high concentrations for full-length antibodies. Therefore, determining the sequence and structural features that differentiate aggregation-resistant antibodies from aggregation-prone ones is critical to improving their activity. We have investigated the molecular origins of antibody aggregation for human V(H) domain antibodies that differ only in the sequence of the loops containing their complementarity determining regions (CDRs), yet such antibodies possess dramatically different aggregation propensities in a manner not correlated with their conformational stabilities. We find the propensity of these antibodies to aggregate after being transiently unfolded is not a distributed property of the CDR loops, but can be localized to aggregation hotspots within and near the first CDR (CDR1). Moreover, we have identified a triad of charged mutations within CDR1 and a single charged mutation adjacent to CDR1 that endow the poorly soluble variant with the desirable biophysical properties of the aggregation-resistant antibody. Importantly, we find that several other charged mutations in CDR1, non-CDR loops and the antibody scaffold are incapable of preventing aggregation. We expect that our identification of aggregation hotspots that govern antibody aggregation within and proximal to CDR loops will guide the design and selection of antibodies that not only possess high affinity and conformational stability, but also extreme resistance to aggregation.

Resveratrol selectively remodels soluble oligomers and fibrils of amyloid Abeta into off-pathway conformers.

Misfolded proteins associated with diverse aggregation disorders assemble not only into a single toxic conformer but rather into a suite of aggregated conformers with unique biochemical properties and toxicities. To what extent small molecules can target and neutralize specific aggregated conformers is poorly understood. Therefore, we have investigated the capacity of resveratrol to recognize and remodel five conformers (monomers, soluble oligomers, non-toxic oligomers, fibrillar intermediates, and amyloid fibrils) of the Abeta1-42 peptide associated with Alzheimer disease. We find that resveratrol selectively remodels three of these conformers (soluble oligomers, fibrillar intermediates, and amyloid fibrils) into an alternative aggregated species that is non-toxic, high molecular weight, and unstructured. Surprisingly, resveratrol does not remodel non-toxic oligomers or accelerate Abeta monomer aggregation despite that both conformers possess random coil secondary structures indistinguishable from soluble oligomers and significantly different from their beta-sheet rich, fibrillar counterparts. We expect that resveratrol and other small molecules with similar conformational specificity will aid in illuminating the conformational epitopes responsible for Abeta-mediated toxicity.