Application of Machine Learning Models to Biomedical and Information System Signals From Critically Ill Adults

BACKGROUND

Machine learning (ML)-derived notifications for impending episodes of hemodynamic instability and respiratory failure events are interesting because they can alert physicians in time to intervene before these complications occur.

RESEARCH QUESTION

Do ML alerts, telemedicine system (TS)-generated alerts, or biomedical monitors (BMs) have superior performance for predicting episodes of intubation or administration of vasopressors?

STUDY DESIGN AND METHODS

An ML algorithm was trained to predict intubation and vasopressor initiation events among critically ill adults. Its performance was compared with BM alarms and TS alerts.

RESULTS

ML notifications were substantially more accurate and precise, with 50-fold lower alarm burden than TS alerts for predicting vasopressor initiation and intubation events. ML notifications of internal validation cohorts demonstrated similar performance for independent academic medical center external validation and COVID-19 cohorts. Characteristics were also measured for a control group of recent patients that validated event detection methods and compared TS alert and BM alarm performance. The TS test characteristics were substantially better, with 10-fold less alarm burden than BM alarms. The accuracy of ML alerts (0.87-0.94) was in the range of other clinically actionable tests; the accuracy of TS (0.28-0.53) and BM (0.019-0.028) alerts were not. Overall test performance (F scores) for ML notifications were more than fivefold higher than for TS alerts, which were higher than those of BM alarms.

INTERPRETATION

ML-derived notifications for clinically actioned hemodynamic instability and respiratory failure events represent an advance because the magnitude of the differences of accuracy, precision, misclassification rate, and pre-event lead time is large enough to allow more proactive care and has markedly lower frequency and interruption of bedside physician work flows.

Validation of an Automatic Tagging System for Identifying Respiratory and Hemodynamic Deterioration Events in the Intensive Care Unit

Objectives

Predictive models for critical events in the intensive care unit (ICU) might help providers anticipate patient deterioration. At the heart of predictive model development lies the ability to accurately label significant events, thereby facilitating the use of machine learning and similar strategies. We conducted this study to establish the validity of an automated system for tagging respiratory and hemodynamic deterioration by comparing automatic tags to tagging by expert reviewers.

Methods

This retrospective cohort study included 72,650 unique patient stays collected from Electronic Medical Records of the University of Massachusetts’ eICU. An enriched subgroup of stays was manually tagged by expert reviewers. The tags generated by the reviewers were compared to those generated by an automated system.

Results

The automated system was able to rapidly and efficiently tag the complete database utilizing available clinical data. The overall agreement rate between the automated system and the clinicians for respiratory and hemodynamic deterioration tags was 89.4% and 87.1%, respectively. The automatic system did not add substantial variability beyond that seen among the reviewers.

Conclusions

We demonstrated that a simple rule-based tagging system could provide a rapid and accurate tool for mass tagging of a compound database. These types of tagging systems may replace human reviewers and save considerable resources when trying to create a validated, labeled database used to train artificial intelligence algorithms. The ability to harness the power of artificial intelligence depends on efficient clinical validation of targeted conditions; hence, these systems and the methodology used to validate them are crucial.

Seasonal Genetic Drift of Human Influenza A Virus Quasispecies Revealed by Deep Sequencing

After a pandemic wave in 2009 following their introduction in the human population, the H1N1pdm09 viruses replaced the previously circulating, pre-pandemic H1N1 virus and, along with H3N2 viruses, are now responsible for the seasonal influenza type A epidemics. So far, the evolutionary potential of influenza viruses has been mainly documented by consensus sequencing data. However, like other RNA viruses, influenza A viruses exist as a population of diverse, albeit related, viruses, or quasispecies. Interest in this quasispecies nature has increased with the development of next generation sequencing (NGS) technologies that allow a more in-depth study of the genetic variability. NGS deep sequencing methodologies were applied to determine the whole genome genetic heterogeneity of the three categories of influenza A viruses that circulated in humans between 2007 and 2012 in France, directly from clinical respiratory specimens. Mutation frequencies and single nucleotide polymorphisms were used for comparisons to address the level of natural intrinsic heterogeneity of influenza A viruses. Clear differences in single nucleotide polymorphism profiles between seasons for a given subtype also revealed the constant genetic drift that human influenza A virus quasispecies undergo.

SMYD1 is the underlying gene for the AnWj-negative blood group phenotype

BACKGROUND

AnWj is a high-incidence blood group antigen associated with three clinical disorders: lymphoid malignancies, immunologic disorders, and autoimmune hemolytic anemia. The aim of this study was to determine the genetic basis of an inherited AnWj-negative phenotype.

METHODS

We identified a consanguineous family with two AnWj-negative siblings and 4 additional AnWj-negative individuals without known familial relationship to the index family. We performed exome sequencing in search for rare homozygous variants shared by the two AnWj-negative siblings of the index family and searched for these variants in the four non-related AnWj-negative individuals.

RESULTS

Exome sequencing revealed seven candidate genes that showed complete segregation in the index family and for which the two AnWj-negative siblings were homozygous. However, the four additional non-related AnWj-negative subjects were homozygous for only one of these variants, rs114851602 (R320Q) in the SMYD1 gene. Considering the frequency of the minor allele, the chance of randomly finding 4 consecutive such individuals is 2.56 × 10^-18 .

CONCLUSION

We present genetic and statistical evidence that the R320Q substitution in SMYD1 underlies an inherited form of the AnWj-negative blood group phenotype. The mechanism by which the mutation leads to this phenotype remains to be determined.

Rare genetic variants in Tunisian Jewish patients suffering from age-related macular degeneration

PURPOSE

To explore the molecular basis of familial, early onset, age-related macular degeneration (AMD) with diverse phenotypes, using whole exome sequencing (WES).

METHODS

We performed WES on four patients (two sibs from two families) manifesting early-onset AMD and searched for disease-causing genetic variants in previously identified macular degeneration related genes. Validation studies of the variants included bioinformatics tools, segregation analysis of mutations within the families and mutation screening in an AMD cohort of patients.

RESULTS

The index patients were in their 50s when diagnosed and displayed a wide variety of clinical AMD presentations: from limited drusen in the posterior pole to multiple basal-laminar drusen extending peripherally. Severe visual impairment due to extensive geographic atrophy and/or choroidal-neovascularisation was common by the age of 75 years. Approximately, 400 000 genomic variants for each DNA sample were included in the downstream bioinformatics analysis, which ended in the discovery of two novel variants; in one family a single bp deletion was identified in the Hemicentin (HMCN1) gene (c.4162delC), whereas in the other, a missense variant (p.V412M) in the Complement Factor-I (CFI) gene was found. Screening for these variants in a cohort of patients with AMD identified another family with the CFI variant.

CONCLUSIONS

This report uses WES to uncover rare genetic variants in AMD. A null-variant in HMCN1 has been identified in one AMD family, and a missense variant in CFI was discovered in two other families. These variants confirm the genetic complexity and significance of rare genetic variants in the pathogenesis of AMD.

Deep sequencing analysis of viral infection and evolution allows rapid and detailed characterization of viral mutant spectrum

Motivation: The study of RNA virus populations is a challenging task. Each population of RNA virus is composed of a collection of different, yet related genomes often referred to as mutant spectra or quasispecies. Virologists using deep sequencing technologies face major obstacles when studying virus population dynamics, both experimentally and in natural settings due to the relatively high error rates of these technologies and the lack of high performance pipelines. In order to overcome these hurdles we developed a computational pipeline, termed ViVan (Viral Variance Analysis). ViVan is a complete pipeline facilitating the identification, characterization and comparison of sequence variance in deep sequenced virus populations.

Results: Applying ViVan on deep sequenced data obtained from samples that were previously characterized by more classical approaches, we uncovered novel and potentially crucial aspects of virus populations. With our experimental work, we illustrate how ViVan can be used for studies ranging from the more practical, detection of resistant mutations and effects of antiviral treatments, to the more theoretical temporal characterization of the population in evolutionary studies.

Availability and implementation: Freely available on the web at http://www.vivanbioinfo.org

Contact: nshomron@post.tau.ac.il

Supplementary information:Supplementary data are available at Bioinformatics online.

Patch Finder Plus (PFplus): a web server for extracting and displaying positive electrostatic patches on protein surfaces

Positively charged electrostatic patches on protein surfaces are usually indicative of nucleic acid binding interfaces. Interestingly, many proteins which are not involved in nucleic acid binding possess large positive patches on their surface as well. In some cases, the positive patches on the protein are related to other functional properties of the protein family. PatchFinderPlus (PFplus) http://pfp.technion.ac.il is a web-based tool for extracting and displaying continuous electrostatic positive patches on protein surfaces. The input required for PFplus is either a four letter PDB code or a protein coordinate file in PDB format, provided by the user. PFplus computes the continuum electrostatics potential and extracts the largest positive patch for each protein chain in the PDB file. The server provides an output file in PDB format including a list of the patch residues. In addition, the largest positive patch is displayed on the server by a graphical viewer (Jmol), using a simple color coding.