Background: Homozygous familial hypercholesterolemia (HoFH) is a rare, treatment‐resistant disorder characterized by early‐onset atherosclerotic and aortic valvular cardiovascular disease if left untreated. Contemporary information on HoFH in the United States is lacking, and the extent of underdiagnosis and undertreatment is uncertain. Methods and Results: Data were analyzed from 67 children and adults with clinically diagnosed HoFH from the CASCADE (Cascade Screening for Awareness and Detection) FH Registry. Genetic diagnosis was confirmed in 43 patients. We used the clinical characteristics of genetically confirmed patients with HoFH to query the Family Heart Database, a US anonymized payer health database, to estimate the number of patients with similar lipid profiles in a “real‐world” setting. Untreated low‐density lipoprotein cholesterol levels were lower in adults than children (533 versus 776 mg/dL; P =0.001). At enrollment, atherosclerotic cardiovascular disease and supravalvular and aortic valve stenosis were present in 78.4% and 43.8% and 25.5% and 18.8% of adults and children, respectively. At most recent follow‐up, despite multiple lipid‐lowering treatment, low‐density lipoprotein cholesterol goals were achieved in only a minority of adults and children. Query of the Family Heart Database identified 277 individuals with profiles similar to patients with genetically confirmed HoFH. Advanced lipid‐lowering treatments were prescribed for 18%; 40% were on no lipid‐lowering treatment; atherosclerotic cardiovascular disease was reported in 20%; familial hypercholesterolemia diagnosis was uncommon. Conclusions: Only patients with the most severe HoFH phenotypes are diagnosed early. HoFH remains challenging to treat. Results from the Family Heart Database indicate HoFH is systemically underdiagnosed and undertreated. Earlier screening, aggressive lipid‐lowering treatments, and guideline implementation are required to reduce disease burden in HoFH.
Machine learning (ML) models trained for triggering clinical decision support (CDS) are typically either accurate or interpretable but not both. Scaling CDS to the panoply of clinical use cases while mitigating risks to patients will require many ML models be intuitively interpretable for clinicians. To this end, we adapted a symbolic regression method, coined the feature engineering automation tool (FEAT), to train concise and accurate models from high-dimensional electronic health record (EHR) data. We first present an in-depth application of FEAT to classify hypertension, hypertension with unexplained hypokalemia, and apparent treatment-resistant hypertension (aTRH) using EHR data for 1200 subjects receiving longitudinal care in a large healthcare system. FEAT models trained to predict phenotypes adjudicated by chart review had equivalent or higher discriminative performance (p < 0.001) and were at least three times smaller (p < 1 × 10−6) than other potentially interpretable models. For aTRH, FEAT generated a six-feature, highly discriminative (positive predictive value = 0.70, sensitivity = 0.62), and clinically intuitive model. To assess the generalizability of the approach, we tested FEAT on 25 benchmark clinical phenotyping tasks using the MIMIC-III critical care database. Under comparable dimensionality constraints, FEAT’s models exhibited higher area under the receiver-operating curve scores than penalized linear models across tasks (p < 6 × 10−6). In summary, FEAT can train EHR prediction models that are both intuitively interpretable and accurate, which should facilitate safe and effective scaling of ML-triggered CDS to the panoply of potential clinical use cases and healthcare practices.
Rationale: Single-nucleotide polymorphisms near the ILRUN (inflammation and lipid regulator with ubiquitin-associated–like and NBR1 [next to BRCA1 gene 1 protein]-like domains) gene are genome-wide significantly associated with plasma lipid traits and coronary artery disease (CAD), but the biological basis of this association is unknown. Objective: To investigate the role of ILRUN in plasma lipid and lipoprotein metabolism. Methods and Results: ILRUN encodes a protein that contains a ubiquitin-associated–like domain, suggesting that it may interact with ubiquitinylated proteins. We generated mice globally deficient for Ilrun and found they had significantly lower plasma cholesterol levels resulting from reduced liver lipoprotein production. Liver transcriptome analysis uncovered altered transcription of genes downstream of lipid-related transcription factors, particularly PPARα (peroxisome proliferator-activated receptor alpha), and livers from Ilrun -deficient mice had increased PPARα protein. Human ILRUN was shown to bind to ubiquitinylated proteins including PPARα, and the ubiquitin-associated–like domain of ILRUN was found to be required for its interaction with PPARα. Conclusions: These findings establish ILRUN as a novel regulator of lipid metabolism that promotes hepatic lipoprotein production. Our results also provide functional evidence that ILRUN may be the casual gene underlying the observed genetic associations with plasma lipids at 6p21 in human.
Angiopoietin-like protein 3 (ANGPTL3) is a hepatically secreted protein and therapeutic target for reducing plasma triglyceride-rich lipoproteins and low-density lipoprotein (LDL) cholesterol. Although ANGPTL3 modulates the metabolism of circulating lipoproteins, its role in triglyceride-rich lipoprotein assembly and secretion remains unknown. CRISPR-associated protein 9 (CRISPR/Cas9) was used to target ANGPTL3 in HepG2 cells (ANGPTL3−/−) whereupon we observed ∼50% reduction of apolipoprotein B100 (ApoB100) secretion, accompanied by an increase in ApoB100 early presecretory degradation via a predominantly lysosomal mechanism. Despite defective particle secretion in ANGPTL3−/− cells, targeted lipidomic analysis did not reveal neutral lipid accumulation in ANGPTL3−/− cells; rather ANGPTL3−/− cells demonstrated decreased secretion of newly synthesized triglycerides and increased fatty acid oxidation. Furthermore, RNA sequencing demonstrated significantly altered expression of key lipid metabolism genes, including targets of peroxisome proliferator-activated receptor α, consistent with decreased lipid anabolism and increased lipid catabolism. In contrast, CRISPR/Cas9 LDL receptor (LDLR) deletion in ANGPTL3−/− cells did not result in a secretion defect at baseline, but proteasomal inhibition strongly induced compensatory late presecretory degradation of ApoB100 and impaired its secretion. Additionally, these ANGPTL3−/−;LDLR−/− cells rescued the deficient LDL clearance of LDLR−/− cells. In summary, ANGPTL3 deficiency in the presence of functional LDLR leads to the production of fewer lipoprotein particles due to early presecretory defects in particle assembly that are associated with adaptive changes in intrahepatic lipid metabolism. In contrast, when LDLR is absent, ANGPTL3 deficiency is associated with late presecretory regulation of ApoB100 degradation without impaired secretion. Our findings therefore suggest an unanticipated intrahepatic role for ANGPTL3, whose function varies with LDLR status.