Genomic prediction using machine learning: a comparison of the performance of regularized regression, ensemble, instance-based and deep learning methods on synthetic and empirical data.

BACKGROUND: The accurate prediction of genomic breeding values is central to genomic selection in both plant and animal breeding studies. Genomic prediction involves the use of thousands of molecular markers spanning the entire genome and therefore requires methods able to efficiently handle high dimensional data. Not surprisingly, machine learning methods are becoming widely advocated for and used in genomic prediction studies. These methods encompass different groups of supervised and unsupervised learning methods. Although several studies have compared the predictive performances of individual methods, studies comparing the predictive performance of different groups of methods are rare. However, such studies are crucial for identifying (i) groups of methods with superior genomic predictive performance and assessing (ii) the merits and demerits of such groups of methods relative to each other and to the established classical methods. Here, we comparatively evaluate the genomic predictive performance and informally assess the computational cost of several groups of supervised machine learning methods, specifically, regularized regression methods, deep, ensemble and instance-based learning algorithms, using one simulated animal breeding dataset and three empirical maize breeding datasets obtained from a commercial breeding program. RESULTS: Our results show that the relative predictive performance and computational expense of the groups of machine learning methods depend upon both the data and target traits and that for classical regularized methods, increasing model complexity can incur huge computational costs but does not necessarily always improve predictive accuracy. Thus, despite their greater complexity and computational burden, neither the adaptive nor the group regularized methods clearly improved upon the results of their simple regularized counterparts. This rules out selection of one procedure among machine learning methods for routine use in genomic prediction. The results also show that, because of their competitive predictive performance, computational efficiency, simplicity and therefore relatively few tuning parameters, the classical linear mixed model and regularized regression methods are likely to remain strong contenders for genomic prediction. CONCLUSIONS: The dependence of predictive performance and computational burden on target datasets and traits call for increasing investments in enhancing the computational efficiency of machine learning algorithms and computing resources.

A Very Rare Case of Coronary Artery Bypass Grafting in a Progeria Child.

Hutchinson-Gilford progeria syndrome is a rare genetic disorder, characterized by progressive premature aging and early death in the first or second decade of life, usually secondary to cardiovascular events (myocardial infarction and stroke). We report a case of a 14-year-old boy with progeria syndrome and cardiac arrest due to myocardial infarction, who was submitted to an immediate coronary angiography which revealed left main stem and three-vessel coronary artery disease. A prompt double bypass coronary artery grafting surgery was performed, and, despite successful coronary reperfusion, the patient remained in coma and brain death was declared on fourth day after surgery.

Troponin I as a marker of right ventricular dysfunction and severity of pulmonary embolism.

BACKGROUND: Cardiac troponin I (cTnI) is a specific marker which allows detection of minor myocardial cell damage. In patients with severe pulmonary embolism (PE), the rise in pulmonary artery pressure can lead to progressive right ventricular dysfunction (RVD), and clinical studies have demonstrated the presence of ischemia and even right ventricular infarction. Our aims were to determine the prevalence and diagnostic utility of cTnI in identifying patients with RVD and to ascertain whether it correlates with severity of PE. METHODS: We studied 77 patients with PE diagnosed by pulmonary angiography, ventilation-perfusion lung scan, spiral computed tomography scan or a combination of abnormal echocardiogram with clinical presentation suggestive of PE or with positive subsidiary exams (d-dimers, venous Doppler of the lower limbs, ECG, blood gas analysis). We further classified the PE according to the European Society of Cardiology severity levels, the PE being: 1) massive, if there was shock and/or hypotension; 2) submassive, if we found right ventricular hypokinesis on the echocardiogram; and 3) non-massive, in the remaining cases. We considered the highest cTnI serum value from the admission to 24 hours and a normal value of < 0.10 ng/ml. RESULTS: Among the 60 patients with cTnI measurements, 42 had elevated values. Among those with RVD, 26 (81.3%) had increased cTnI levels and only 14 (35%) with elevated cTnI values did not have RVD, indicating that positive cTnI tests were significantly associated with RVD (p = 0.038). Patients with positive cTnI tests had earlier onset of symptoms (24.0 vs. 144.0 hours, p=0.02), higher prevalence of emboli in proximal vessels (pulmonary trunk and right or left main pulmonary arteries) (OR = 12, CI= 1.6-88.7), and received more thrombolytic therapy (OR = 5.4, CI = 1.1-26.8) than those with normal cTnI tests. cTnI levels were higher among patients with submassive PE (median: 0.77 ng/ml) and lower in those with non-massive PE (0.08 mg/ml, p < 0.05). CONCLUSIONS: Around 70% of patients with PE have elevated cTnI values and this test is significantly associated with RVD. cTnI measurements provide additional information in the evaluation of patients with PE by identifying more severe cases and those at increased risk of hemodynamic deterioration, who can benefit from more aggressive therapeutic strategies.