tRNAscan-SE 2.0 improves tRNA detection and classification by using a larger training set and specialized models for different tRNA isotypes and clades. It employs a new comparative multi-model strategy, allowing functional prediction based on both the anticodon and the highest-scoring isotype model. This enhances the program's ability to detect tRNA-derived SINEs and other pseudogenes. For the first time, tRNAscan-SE 2.0 includes fast and accurate detection of mitochondrial tRNAs. Overall, tRNA detection sensitivity and specificity are improved for all isotypes, particularly those with specialized models for selenocysteine and the three subtypes of tRNA genes encoding a CAU anticodon. These enhancements provide more accurate and detailed tRNA annotations and may highlight tRNAs with novel traits. tRNAscan-SE 2.0 uses Infernal 1.1 for sequence searches, incorporating new covariance models for better functional classification. It includes a new 'high confidence' filter to distinguish tRNA-derived repetitive elements from canonical tRNAs. The program also allows for the detection of noncanonical introns in archaeal tRNA genes and accurate identification of vertebrate mitochondrial tRNAs. The default score cutoff remains at 20 bits, with a higher threshold of 30 bits for more stringent searches. tRNAscan-SE 2.0 improves performance by using domain-specific and isotype-specific covariance models. It includes models for selenocysteine tRNAs and isoleucine tRNAs decoding AUA. The program also includes models for noncanonical introns in archaeal tRNAs and mitochondrial tRNAs in vertebrates. It compares predictions with NCBI RefSeq annotations to evaluate accuracy and performance. tRNAscan-SE 2.0 shows improved sensitivity and accuracy in detecting tRNA genes across various organisms. It correctly identifies more tRNA genes in bacterial and archaeal genomes, including selenocysteine tRNAs. In eukaryotic genomes, it shows similar performance to tRNAscan-SE 1.3, with a slight improvement in detection of low-scoring tRNA genes. The program also includes a post-scan filter to distinguish tRNA-derived repetitive elements from canonical tRNAs. tRNAscan-SE 2.0 is slightly slower than tRNAscan-SE 1.3 for organisms with small genomes but much faster for large eukaryotic genomes. It has a lower false positive rate than the original version, with an estimated selectivity of 0.00005 per million bp. The program is effective in distinguishing between initiator methionine, elongator methionine, and isoleucine 2 tRNAs, which share the same anticodon.tRNAscan-SE 2.0 improves tRNA detection and classification by using a larger training set and specialized models for different tRNA isotypes and clades. It employs a new comparative multi-model strategy, allowing functional prediction based on both the anticodon and the highest-scoring isotype model. This enhances the program's ability to detect tRNA-derived SINEs and other pseudogenes. For the first time, tRNAscan-SE 2.0 includes fast and accurate detection of mitochondrial tRNAs. Overall, tRNA detection sensitivity and specificity are improved for all isotypes, particularly those with specialized models for selenocysteine and the three subtypes of tRNA genes encoding a CAU anticodon. These enhancements provide more accurate and detailed tRNA annotations and may highlight tRNAs with novel traits. tRNAscan-SE 2.0 uses Infernal 1.1 for sequence searches, incorporating new covariance models for better functional classification. It includes a new 'high confidence' filter to distinguish tRNA-derived repetitive elements from canonical tRNAs. The program also allows for the detection of noncanonical introns in archaeal tRNA genes and accurate identification of vertebrate mitochondrial tRNAs. The default score cutoff remains at 20 bits, with a higher threshold of 30 bits for more stringent searches. tRNAscan-SE 2.0 improves performance by using domain-specific and isotype-specific covariance models. It includes models for selenocysteine tRNAs and isoleucine tRNAs decoding AUA. The program also includes models for noncanonical introns in archaeal tRNAs and mitochondrial tRNAs in vertebrates. It compares predictions with NCBI RefSeq annotations to evaluate accuracy and performance. tRNAscan-SE 2.0 shows improved sensitivity and accuracy in detecting tRNA genes across various organisms. It correctly identifies more tRNA genes in bacterial and archaeal genomes, including selenocysteine tRNAs. In eukaryotic genomes, it shows similar performance to tRNAscan-SE 1.3, with a slight improvement in detection of low-scoring tRNA genes. The program also includes a post-scan filter to distinguish tRNA-derived repetitive elements from canonical tRNAs. tRNAscan-SE 2.0 is slightly slower than tRNAscan-SE 1.3 for organisms with small genomes but much faster for large eukaryotic genomes. It has a lower false positive rate than the original version, with an estimated selectivity of 0.00005 per million bp. The program is effective in distinguishing between initiator methionine, elongator methionine, and isoleucine 2 tRNAs, which share the same anticodon.