Predicting Mitochondrial tRNA Modification

M itochondria are integral to proper cell function, and mutations in its small genome (mtDNA) are associated with many diseases, along with the progression of normal aging [9]. While mtDNA has been extensively studied, not much is known about transcriptional variations of mitochondrial genes [3]. Recently, tRNA modifications have been the focus of intense study owing to their putative role in diseases [4]. Identifying genes responsible for mitochondrial tRNA modification is an important step towards a better understanding of transcriptional variation. To the best of my knowledge, a machine learning approach has never been utilized in order to identify such genes. Here, I used penalized linear regression to predict tRNA modification activity using gene expression as features. After controlling for confounding factors, I was able to predict modification activity at several putative modified tRNA positions. These models explained between 19% and 51% of the variance. Most notably, I used preconditioned lasso [7] which yielded four promising gene candidates that affect tRNA modification: ALKBH8 (an E. coli homolog DNA repair enzyme), MAD2L1 Binding Protein, C1orf103 (an open reading frame gene with unknown function), and TRMT5 (tRNA methyltransferase 5).