An electrocardiogram-based AI algorithm for early detection of pulmonary hypertension.

BACKGROUND: Early diagnosis of pulmonary hypertension (PH) is critical for effective treatment and management. We aimed to develop and externally validate an artificial intelligence algorithm that could serve as a PH screening tool, based on analysis of a standard 12-lead ECG. METHODS: The PH Early Detection Algorithm (PH-EDA) is a convolutional neural network developed using retrospective ECG voltage-time data, with patients classified as "PH-likely" or "PH-unlikely" (controls) based on right heart catheterisation or echocardiography. In total, 39 823 PH-likely patients and 219 404 control patients from Mayo Clinic were randomly split into training (48%), validation (12%) and test (40%) sets. ECGs taken within 1 month of PH diagnosis (diagnostic dataset) were used to train the PH-EDA at Mayo Clinic. Performance was tested on diagnostic ECGs within the test sets from Mayo Clinic (n=16 175/87 998 PH-likely/controls) and Vanderbilt University Medical Center (VUMC; n=6045/24 256 PH-likely/controls). In addition, performance was tested on ECGs taken 6-18 months (pre-emptive dataset), and up to 5 years prior to a PH diagnosis at both sites. RESULTS: Performance testing yielded an area under the receiver operating characteristic curve (AUC) of 0.92 and 0.88 in the diagnostic test sets at Mayo Clinic and VUMC, respectively, and 0.86 and 0.81, respectively, in the pre-emptive test sets. The AUC remained a minimum of 0.79 at Mayo Clinic and 0.73 at VUMC up to 5 years before diagnosis. CONCLUSION: The PH-EDA can detect PH at diagnosis and 6-18 months prior, demonstrating the potential to accelerate diagnosis and management of this debilitating disease.

Development and evaluation of a predictive algorithm for unsatisfactory response among patients with pulmonary arterial hypertension using health insurance claims data.

OBJECTIVE: This study aimed to develop and validate a predictive algorithm for unsatisfactory response to initial pulmonary arterial hypertension (PAH) therapy using health insurance claims. METHODS: Adult patients with PAH initiated on a first PAH therapy (index date) were identified from Optum's de-identified Clinformatics Data Mart Database (1/1/2010-12/31/2019). A random survival forest algorithm was developed using patient-month data and predicted the "survival function" (i.e. risk of not having unsatisfactory response) over time. For each patient-month observation, risk factors were assessed in the 12 months prior. Unsatisfactory response was defined as the first instance of (1) new PAH therapy, (2) PAH-related hospitalization or emergency room visit, (3) lung transplant or atrial septostomy, (4) PAH-related death or (5) chronic oxygen therapy initiation. To facilitate use in clinical practice, a simplified risk score was also developed based on a linear combination of the most important risk factors identified in the algorithm. RESULTS: In total, 4781 patients were included (median age = 69.0 years; 58.6% female). Over a median follow-up of 14.0 months, 3169 (66.3%) had an unsatisfactory response. The most important risk factors included in the algorithm were healthcare resource use (i.e. PAH-related outpatient visits, pulmonologist visits, cardiologist visits, all-cause hospitalizations), time since first PAH diagnosis, time since index date, Charlson Comorbidity Index, dyspnea, and age. Predictive accuracy was good for the full algorithm (C-statistic: 0.732) but was slightly lower for the simplified risk score (C-statistic: 0.668). CONCLUSION: The present claims-based algorithm performed well in predicting time to unsatisfactory response following initial PAH therapy.

Burden of pulmonary hypertension in patients with portal hypertension in the United States: a retrospective database study.

Patients with portal hypertension may develop pulmonary hypertension. The economic implications of these comorbidities have not been systematically assessed. We compared healthcare resource utilization and costs in the United States between patients with co-existing portal hypertension and pulmonary hypertension (pulmonary hypertension cohort) and a matched cohort of portal hypertension patients without pulmonary hypertension (control cohort). In this retrospective analysis, adult pulmonary hypertension and control patients were identified from the Optum® Clinformatics® Data Mart database between 1 July 2014 and 30 June 2018. All patients had ≥2 claims with diagnosis codes for portal hypertension; pulmonary hypertension patients had ≥2 claims with diagnosis codes for pulmonary hypertension; controls could not have pulmonary hypertension diagnoses or any claims for pulmonary arterial hypertension-specific medications. Controls were matched to pulmonary hypertension patients by age, sex, Charlson comorbidity index score, and liver diseases. We assessed 12-month healthcare resource utilization and costs. Each cohort included 146 patients. During follow-up, pulmonary hypertension cohort patients were more likely than controls to experience a hospitalization (51% vs. 32%, P = 0.0014) and an emergency room visit (55% vs. 41%, P = 0.026). The average annual total cost was higher in pulmonary hypertension patients than for matched controls ($119,912 vs. $81,839, P < 0.0001). After covariate adjustment, costs for pulmonary hypertension cohort patients were 1.47 times higher than those for controls (P = 0.0197). These findings suggest that patients with portal hypertension and co-existing pulmonary hypertension are at a greater risk for hospitalization and incur higher mean annual total costs than portal hypertension patients without pulmonary hypertension.

Development of a genetic system for the chemolithoautotrophic bacterium Thiobacillus denitrificans.

Thiobacillus denitrificans is a widespread, chemolithoautotrophic bacterium with an unusual and environmentally relevant metabolic repertoire, which includes its ability to couple denitrification to sulfur compound oxidation; to catalyze anaerobic, nitrate-dependent oxidation of Fe(II) and U(IV); and to oxidize mineral electron donors. Recent analysis of its genome sequence also revealed the presence of genes encoding two [NiFe]hydrogenases, whose role in metabolism is unclear, as the sequenced strain does not appear to be able to grow on hydrogen as a sole electron donor under denitrifying conditions. In this study, we report the development of a genetic system for T. denitrificans, with which insertion mutations can be introduced by homologous recombination and complemented in trans. The antibiotic sensitivity of T. denitrificans was characterized, and a procedure for transformation with foreign DNA by electroporation was established. Insertion mutations were generated by in vitro transposition, the mutated genes were amplified by the PCR, and the amplicons were introduced into T. denitrificans by electroporation. The IncP plasmid pRR10 was found to be a useful vector for complementation. The effectiveness of the genetic system was demonstrated with the hynL gene, which encodes the large subunit of a [NiFe]hydrogenase. Interruption of hynL in a hynL::kan mutant resulted in a 75% decrease in specific hydrogenase activity relative to the wild type, whereas complementation of the hynL mutation resulted in activity that was 50% greater than that of the wild type. The availability of a genetic system in T. denitrificans will facilitate our understanding of the genetics and biochemistry underlying its unusual metabolism.

Genome differences that distinguish Bacillus anthracis from Bacillus cereus and Bacillus thuringiensis.

The three species of the group 1 bacilli, Bacillus anthracis, B. cereus, and B. thuringiensis, are genetically very closely related. All inhabit soil habitats but exhibit different phenotypes. B. anthracis is the causative agent of anthrax and is phylogenetically monomorphic, while B. cereus and B. thuringiensis are genetically more diverse. An amplified fragment length polymorphism analysis described here demonstrates genetic diversity among a collection of non-anthrax-causing Bacillus species, some of which show significant similarity to B. anthracis. Suppression subtractive hybridization was then used to characterize the genomic differences that distinguish three of the non-anthrax-causing bacilli from B. anthracis Ames. Ninety-three DNA sequences that were present in B. anthracis but absent from the non-anthrax-causing Bacillus genomes were isolated. Furthermore, 28 of these sequences were not found in a collection of 10 non-anthrax-causing Bacillus species but were present in all members of a representative collection of B. anthracis strains. These sequences map to distinct loci on the B. anthracis genome and can be assayed simultaneously in multiplex PCR assays for rapid and highly specific DNA-based detection of B. anthracis.

Establishment of uncharacterized plasmids in Escherichia coli by in vitro transposition.

We present a simple approach that permits any circular plasmid, such as uncharacterized plasmids from diverse prokaryotes, to be established in Escherichia coli, thereby facilitating subsequent structural and functional studies. An in vitro transposition reaction is used to introduce a well-characterized replicon and selectable marker into purified plasmids, which are then used to transform E. coli. The approach was demonstrated using a small 3.4-kb archaeal plasmid and a large 60-kb uncharacterized plasmid from a Gram-negative bacterium. Replicon function in E. coli was tested for each plasmid, and direct sequencing of the large plasmid revealed similarity to restriction-modification systems.

Genome plasticity in Yersinia pestis.

Yersinia pestis, the causative agent of bubonic plague, emerged recently (<20000 years ago) as a clone of Yersinia pseudotuberculosis. There is scant evidence of genome diversity in Y. pestis, although it is possible to differentiate three biovars (antiqua, mediaevalis or orientalis) based on two biochemical tests. There are a few examples of restriction fragment length polymorphisms (RFLPs) within Y. pestis; however, their genetic basis is poorly understood. In this study, six difference regions (DFRs) were identified in Y. pestis, by using subtractive hybridization, which ranged from 4.6 to 19 kb in size. Four of the DFRs are flanked by insertion sequences, and their sequences show similarity to bacterial genes encoding proteins for flagellar synthesis, ABC transport, insect toxicity and bacteriophage functions. The presence or absence of these DFRs (termed the DFR profile) was demonstrated in 78 geographically diverse strains of Y. pestis. Significant genome plasticity was observed among these strains and suggests the acquisition and deletion of these DNA regions during the recent evolution of Y. pestis. Y. pestis biovar orientalis possesses DFR profiles that are different from antiqua and mediaevalis biovars, reflecting the recent origins of this biovar. Whereas some DFR profiles are specific for antiqua and mediaevalis, some DFR profiles are shared by both biovars. Furthermore, the progenitor of Y. pestis, Y. pseudotuberculosis (an enteric pathogen), possesses its own DFR profile. The DFR profiles detailed here demonstrate genome plasticity within Y. pestis, and they imply evolutionary relationships among the three biovars of Y. pestis, as well as between Y. pestis and Y. pseudotuberculosis.

Use of subtractive hybridization for comprehensive surveys of prokaryotic genome differences.

Comparative bacterial genomics shows that even different isolates of the same bacterial species can vary significantly in gene content. An effective means to survey differences across whole genomes would be highly advantageous for understanding this variation. Here we show that suppression subtractive hybridization (SSH) provides high, representative coverage of regions that differ between similar genomes. Using Helicobacter pylori strains 26695 and J99 as a model, SSH identified approximately 95% of the unique open reading frames in each strain, showing that the approach is effective. Furthermore, combining data from parallel SSH experiments using different restriction enzymes significantly increased coverage compared to using a single enzyme. These results suggest a powerful approach for assessing genome differences among closely related strains when one member of the group has been completely sequenced.