Graph Neural Networks (GNNs) are a popular machine learning framework for solving various graph processing applications. This framework exploits both the graph topology and the feature vectors of the nodes. One of the important applications of GNN is in the semi-supervised node classification task. The accuracy of the node classification using GNN depends on (i) the number and (ii) the choice of the training nodes. In this article, we demonstrate that increasing the training nodes by selecting nodes from the same class that are spread out across non-contiguous subgraphs, can significantly improve the accuracy. We accomplish this by presenting a novel input intervention technique that can be used in conjunction with different GNN classification methods to increase the non-contiguous training nodes and, thereby, improve the accuracy. We also present an output intervention technique to identify misclassified nodes and relabel them with their potentially correct labels. We demonstrate on real-world networks that our proposed methods, both individually and collectively, significantly improve the accuracy in comparison to the baseline GNN algorithms. Both our methods are agnostic. Apart from the initial set of training nodes generated by the baseline GNN methods, our techniques do not need any other extra knowledge about the classes of the nodes. Thus, our methods are modular and can be used as pre-and post-processing steps with many of the currently available GNN methods to improve their accuracy.

Primary mitochondrial disorders are clinically and genetically heterogeneous and remain underdiagnosed in resource-limited settings. We performed a retrospective observational study (March 2016- January 2024) at a tertiary neurology center in Eastern India to characterize the clinical, biochemical, neuroimaging, electrophysiological, and molecular features of suspected mitochondrial disease and to explore interpretable machine-learning approaches for syndromic stratification. Forty-eight patients from 42 unrelated families were classified as mitochondrial encephalomyopathy, lactic acidosis, and stroke-like episodes (MELAS; n = 17), chronic progressive external ophthalmoplegia (CPEO; n = 14), Leber hereditary optic neuropathy (LHON; n = 10), or Leigh syndrome (n = 7). Mean age at presentation was 23.9 years (range: 9 months-60 years), with a slight male predominance. Neuroimaging was abnormal in 23/48 (47.9%) and showed syndrome-concordant patterns, including stroke-like cortical lesions in MELAS and symmetric basal ganglia involvement in Leigh syndrome; brain magnetic resonance imaging was typically normal in CPEO. Elevated blood and/or cerebrospinal fluid lactate was common, and electroencephalographic abnormalities were concentrated in MELAS and Leigh syndrome. Targeted molecular testing in a subset identified pathogenic mtDNA variants consistent with phenotype, including MT-TL1 variants in MELAS, m.11778G>A in MT-ND4 in LHON, and m.8993T>G in MT-ATP6 in Leigh syndrome; no mtDNA deletions were detected in tested CPEO cases. Decision tree and random forest models highlighted clinically intuitive discriminators (e.g., visual loss, external ophthalmoplegia/ptosis, and seizure phenotype), supporting their potential role as transparent triage tools for targeted molecular evaluation. This cohort provides the first detailed characterization of mitochondrial syndromes in Eastern India and supports a pragmatic diagnostic framework integrating bedside phenotyping, targeted assays, and interpretable machine learning.