Prediction of Influenza Complications: Development and Validation of a Machine Learning Prediction Model to Improve and Expand the Identification of Vaccine-Hesitant Patients at Risk of Severe Influenza Complications.

Influenza vaccinations are recommended for high-risk individuals, but few population-based strategies exist to identify individual risks. Patient-level data from unvaccinated individuals, stratified into retrospective cases (n = 111,022) and controls (n = 2,207,714), informed a machine learning model designed to create an influenza risk score; the model was called the Geisinger Flu-Complications Flag (GFlu-CxFlag). The flag was created and validated on a cohort of 604,389 unique individuals. Risk scores were generated for influenza cases; the complication rate for individuals without influenza was estimated to adjust for unrelated complications. Shapley values were used to examine the model's correctness and demonstrate its dependence on different features. Bias was assessed for race and sex. Inverse propensity weighting was used in the derivation stage to correct for biases. The GFlu-CxFlag model was compared to the pre-existing Medial EarlySign Flu Algomarker and existing risk guidelines that describe high-risk patients who would benefit from influenza vaccination. The GFlu-CxFlag outperformed other traditional risk-based models; the area under curve (AUC) was 0.786 [0.783−0.789], compared with 0.694 [0.690−0.698] (p-value < 0.00001). The presence of acute and chronic respiratory diseases, age, and previous emergency department visits contributed most to the GFlu-CxFlag model's prediction. When higher numerical scores were assigned to more severe complications, the GFlu-CxFlag AUC increased to 0.828 [0.823−0.833], with excellent discrimination in the final model used to perform the risk stratification of the population. The GFlu-CxFlag can better identify high-risk individuals than existing models based on vaccination guidelines, thus creating a population-based risk stratification for individual risk assessment and deployment in vaccine hesitancy reduction programs in our health system.

Prediction of progression from pre-diabetes to diabetes: Development and validation of a machine learning model.

AIMS: Identification, a priori, of those at high risk of progression from pre-diabetes to diabetes may enable targeted delivery of interventional programmes while avoiding the burden of prevention and treatment in those at low risk. We studied whether the use of a machine-learning model can improve the prediction of incident diabetes utilizing patient data from electronic medical records. METHODS: A machine-learning model predicting the progression from pre-diabetes to diabetes was developed using a gradient boosted trees model. The model was trained on data from The Health Improvement Network (THIN) database cohort, internally validated on THIN data not used for training, and externally validated on the Canadian AppleTree and the Israeli Maccabi Health Services (MHS) data sets. The model's predictive ability was compared with that of a logistic-regression model within each data set. RESULTS: A cohort of 852 454 individuals with pre-diabetes (glucose ≥ 100 mg/dL and/or HbA1c ≥ 5.7) was used for model training including 4.9 million time points using 900 features. The full model was eventually implemented using 69 variables, generated from 11 basic signals. The machine-learning model demonstrated superiority over the logistic-regression model, which was maintained at all sensitivity levels - comparing AUC [95% CI] between the models; in the THIN data set (0.865 [0.860,0.869] vs 0.778 [0.773,0.784] P < .05), the AppleTree data set (0.907 [0.896, 0.919] vs 0.880 [0.867, 0.894] P < .05) and the MHS data set (0.925 [0.923, 0.927] vs 0.876 [0.872, 0.879] P < .05). CONCLUSIONS: Machine-learning models preserve their performance across populations in diabetes prediction, and can be integrated into large clinical systems, leading to judicious selection of persons for interventional programmes.

Systematic identification of abundant A-to-I editing sites in the human transcriptome.

RNA editing by members of the ADAR (adenosine deaminases acting on RNA) family leads to site-specific conversion of adenosine to inosine (A-to-I) in precursor messenger RNAs. Editing by ADARs is believed to occur in all metazoa, and is essential for mammalian development. Currently, only a limited number of human ADAR substrates are known, whereas indirect evidence suggests a substantial fraction of all pre-mRNAs being affected. Here we describe a computational search for ADAR editing sites in the human transcriptome, using millions of available expressed sequences. We mapped 12,723 A-to-I editing sites in 1,637 different genes, with an estimated accuracy of 95%, raising the number of known editing sites by two orders of magnitude. We experimentally validated our method by verifying the occurrence of editing in 26 novel substrates. A-to-I editing in humans primarily occurs in noncoding regions of the RNA, typically in Alu repeats. Analysis of the large set of editing sites indicates the role of editing in controlling dsRNA stability.

Widespread occurrence of antisense transcription in the human genome.

An increasing number of eukaryotic genes are being found to have naturally occurring antisense transcripts. Here we study the extent of antisense transcription in the human genome by analyzing the public databases of expressed sequences using a set of computational tools designed to identify sense-antisense transcriptional units on opposite DNA strands of the same genomic locus. The resulting data set of 2,667 sense-antisense pairs was evaluated by microarrays containing strand-specific oligonucleotide probes derived from the region of overlap. Verification of specific cases by northern blot analysis with strand-specific riboprobes proved transcription from both DNA strands. We conclude that > or =60% of this data set, or approximately 1,600 predicted sense-antisense transcriptional units, are transcribed from both DNA strands. This indicates that the occurrence of antisense transcription, usually regarded as infrequent, is a very common phenomenon in the human genome. Therefore, antisense modulation of gene expression in human cells may be a common regulatory mechanism.

Large-scale protein annotation through gene ontology.

Recent progress in genomic sequencing, computational biology, and ontology development has presented an opportunity to investigate biological systems from a unique perspective, that is, examining genomes and transcriptomes through the multiple and hierarchical structure of Gene Ontology (GO). We report here our development of GO Engine, a computational platform for GO annotation, and analysis of the resultant GO annotations of human proteins. Protein annotation was centered on sequence homology with GO-annotated proteins and protein domain analysis. Text information analysis and a multiparameter cellular localization predictive tool were also used to increase the annotation accuracy, and to predict novel annotations. The majority of proteins corresponding to full-length mRNA in GenBank, and the majority of proteins in the NR database (nonredundant database of proteins) were annotated with one or more GO nodes in each of the three GO categories. The annotations of GenBank and SWISS-PROT proteins are available to the public at the GO Consortium web site.