tRNAscan-SE 2.0: improved detection and functional classification of transfer RNA genes

Abstract

tRNAscan-SE has been widely used for transfer RNA (tRNA) gene prediction for over twenty years, developed just as the first genomes were decoded. With the massive increase in quantity and phylogenetic diversity of genomes, the accurate detection and functional prediction of tRNAs has become more challenging. Utilizing a vastly larger training set, we created nearly one hundred specialized isotype- and clade-specific models, greatly improving tRNAscan-SE's ability to identify and classify both typical and atypical tRNAs. We employ a new comparative multi-model strategy where predicted tRNAs are scored against a full set of isotype-specific covariance models, allowing functional prediction based on both the anticodon and the highest-scoring isotype model. Comparative model scoring has also enhanced the program's ability to detect tRNA-derived SINEs and other likely pseudogenes. For the first time, tRNAscan-SE also includes fast and highly accurate detection of mitochondrial tRNAs using newly developed models. Overall, tRNA detection sensitivity and specificity is improved for all isotypes, particularly those utilizing specialized models for selenocysteine and the three subtypes of tRNA genes encoding a CAU anticodon. These enhancements will provide researchers with more accurate and detailed tRNA annotation for a wider variety of tRNAs, and may direct attention to tRNAs with novel traits.

Figure 1.

Schematic diagram of tRNAscan-SE 2.0 search algorithm. Three pathways were developed for cytosolic tRNA search modes with the addition of the mitochondrial tRNA search mode. The default method employs Infernal 1.1 (20) with newly built covariance models for similarity search while the legacy search remains the same as tRNAscan-SE 1.3 (1) for backward compatibility.

Figure 2.

Score distributions of tRNA genes comparing tRNAscan-SE 2.0 to 1.3 predictions show substantial agreement, with most differences among low-scoring (<50 bit) tRNA genes. (A) 10 354 tRNA predictions in 217 archaeal genomes, (B) 246 271 predictions in 4036 bacterial genomes and (C) 80 824 predictions in 432 fungal genomes were detected using tRNAscan-SE 2.0 with default search modes for each respective domain. Prediction results were grouped into four categories: (i) ‘Consistent’ between versions, (ii) ‘Isotype Mismatch’ where v2.0 and v1.3 predictions have different isotype classifications, (iii) ‘Novel’ for predictions new in v2.0 and (iv) ‘Not detected’ for prior 1.3 predictions no longer detected by v2.0. Each histogram represents the prediction scores binned by 10 bits in the corresponding comparison category. Vertical black dashed lines represent the median score in the category, and vertical red dashed lines demarcate the 50-bit ‘low score’ threshold.

Figure 3.

Direct comparison of isotype-specific covariance model scores provides an effective classifier of tRNAs with anticodon CAU. Dots represent individual tRNA genes with the highest scores from the initiator methionine/N-formylmethionine model (tRNA^iMet/fMet, red), elongator methionine model (tRNA^Met, blue) or isoleucine-decoding CAU model (tRNA^Ile2, green). Each tRNA gene was scanned with the isotype-specific covariance models of the corresponding genome's domain (A) eukaryotes, (B) bacteria, or (C) archaea.

Figure 4.

Isotype uncertainty in human tRNA-Leu-CAA-5–1. The primary sequence and the secondary structure comparison of (A) tRNA-Met-CAT-3–1, (B) tRNA-Leu-CAA-5–1, and (C) tRNA-Leu-CAA-1–1 shows that tRNA-Leu-CAA-5–1 is more similar to a tRNA^Met than a tRNA^Leu even though it has an anticodon that would decode leucine. The bases highlighted in orange represent the differences between the respective sequence and tRNA-Leu-CAA-5–1. The numbers next to the bases represent the Sprinzl canonical tRNA positions (16).