Diabetic sensorimotor polyneuropathy (DSPN) is a major complication in patients with diabetes mellitus (DM), and early detection or prediction of DSPN is important for preventing or managing neuropathic pain and foot ulcer. Our aim is to delineate whether machine learning techniques are more useful than traditional statistical methods for predicting DSPN in DM patients. Four hundred seventy DM patients were classified into four groups (normal, possible, probable, and confirmed) based on clinical and electrophysiological findings of suspected DSPN. Three ML methods, XGBoost (XGB), support vector machine (SVM), and random forest (RF), and their combinations were used for analysis. RF showed the best area under the receiver operator characteristic curve (AUC, 0.8250) for differentiating between two categories—criteria by clinical findings (normal, possible, and probable groups) and those by electrophysiological findings (confirmed group)—and the result was superior to that of linear regression analysis (AUC = 0.6620). Average values of serum glucose, International Federation of Clinical Chemistry (IFCC), HbA1c, and albumin levels were identified as the four most important predictors of DSPN. In conclusion, machine learning techniques, especially RF, can predict DSPN in DM patients effectively, and electrophysiological analysis is important for identifying DSPN.
1 January 2010 in Journal of Korean Medical Science
More than 20 different RFLPs have been described in the apo AI-CIII-AVI gene cluster and their association with several dyslipidemias has been claimed in different populations. So far, however, the results of their association have been largely inconclusive. We analyzed six previously described RFLPs (Paul-Hayase et al., 1992) in the apo AI-CII-AIV gene cluster by using PCR-amplified DNAs in 132 healthy subjects to document the relationship between the RFLPs and dyslipoproteinemias. The polymorphic sites were located in the promoter region of the apo AI gene (G to A substitution, MspI), in the 3' flanking region of the apo AI gene (PstI), in the 3' noncoding region of the apo CIII gene (SstI), in the first intron of the apo CIII gene (PvuIIa), in the intergenic region of the apo CIII and apo AIV genes (PvuIIb), and in the second intron of the apo AIV gene (XbaI), respectively. The allele frequencies of PstI polymorphism showed no ethnic difference. However, in G to A substitution sites (MspI), Koreans had a lower G allele frequency than that of Caucasians. Among Oriental ethnic groups there were no significant differences in the allele frequencies of SstI and PvuIIa RFLPs, whereas significant differences existed between individuals of Oriental, Caucasian, and/or African origins. Koreans were monomorphic for PvuIIb and XbaI restriction sites. Pairwise linkage disequilibrium analysis revealed two pairs of significant negative linkage disequilibriums (MspI-SstI and SstI-PvuIIa). Although there were no statistical significances, the S2 allele of SstI polymorphism and the G allele of MspI polymorphism seemed to be associated with hypertriglyceridemia. Thus linkage of the G allele of MspI polymorphism and the S2 allele of SstI polymorphism could give an additive effect on an increment of triglyceride levels. We did not find any association between the SstI polymorphism and apo CIII levels. In conclusion, there were no statistically significant relationships detected between the apo AI-CIII-AVI gene cluster polymorphisms and the levels of various lipid parameters.