On 2016 Feb 11, Todd Lowe commented:

We thank Isidore Rigoutsos and co-authors for the helpful, detailed comments. We have addressed all the issues in their commentary.

In brief:

1) The legacy names have been fixed and updated.

2) For now, we have put the pseudogene designations back into the database — this will likely change in the future as we develop a more objective, less arbitrary cutoff for what is called a tRNA pseudogene. We will post release notes when we do make these types of changes in the future, and welcome feedback from users.

3) The name mismatches between the GtRNAdb, NCBI, and HGNC for human tRNAs is due to an incomplete update of HGNC records requested previously, but we are coordinating with them to get these updated quickly.

4) We have put a disclaimer on the front page of the GtRNAdb letting users know that this database is a reflection of what we believe is the state of the art in tRNA detection and annotation. Our own understanding of tRNAs has been greatly enhanced by the use of the new isotype-specific models and increased sensitivity from tRNAscan-SE 2.0 (manuscript in preparation), so the criteria that we previously applied are, in some cases, clearly outdated and can be improved. A number of manuscripts are in preparation detailing these new insights based on our improved tRNA detection methods.

5) If users prefer a static, historical view of tRNAscan-SE gene calls, the prior GtRNAdb will still be available for reference at http://gtrnadb2009.ucsc.edu/ for the foreseeable future.

6) Some endpoints have indeed changed slightly (generally by 3-10 nucleotides total, for just 15 of 600+ total genes) because low-scoring tRNAs are aligned & scored slightly differently by Infernal 1.1 and the new tRNA covariance models, compared to the older software. We have over 60 new isotype-specific covariance models that we are still improving and refining, so small adjustments are expected over the next few months.

7) Of the "new" tRNAs in the GtRNAdb from the human genome, the vast majority are either very low scoring (i.e., they were just below the 20.0 bit covariance model score reporting cutoff used by the old version of tRNAscan-SE; with Infernal and new models, those scores may shift by 2-5 bits, bringing some above the reporting threshold, and some dropping below it). Because these differences are only affecting very low-scoring tRNAs, they appear to have little to no effect on the high-scoring tRNAs used in translation. Also, a fair number appear to be mitochondrial-derived tRNA genes, which tRNAscan-SE 2.0 is now able to detect with high sensitivity. These nuclear-encoded mitochondrial tRNA genes are now recognized to be commonly found in nuclear genomes, just as many mitochondrial proteins have migrated to the nuclear genome.

We regret any inconvenience these issues have caused users in the first few weeks since the database went live in December 2015. We encourage users to email us directly if they have questions or issues we can address. We are working hard to make this a useful, powerful resource to support the rapidly growing field of tRNA research.

Todd M. Lowe, Biomolecular Engineering, University of California, Santa Cruz

On 2016 Jan 21, Isidore Rigoutsos commented:

Comments on the contents of release v2.0 of gtRNAdb

The gtRNAdb repository is a great resource for researchers studying transfer RNAs (tRNAs). With the increasing interest in the roles tRNAs and tRNA fragments play outside the confines of codon translation into amino acids, databases such as gtRNAdb are expected to provide invaluable reference information.

We focused on the Homo sapiens portion of gtRNAdb v2.0 and found the following:

Sweeping, undocumented changes in v2.0. In the above article, the authors state that they populated gtRNAdb v2.0 using a method that underwent major development but has not yet been peer-reviewed (the method is cited as “Chan et al., in preparation”). Specifically, for Homo sapiens the new method resulted in changes affecting 28% of the human tRNA records; the changes comprise

• the deletion of 74 tRNA records originally in v1.0

• the “elevation” of 41 pseudo-tRNAs of v1.0 to tRNAs in v2.0

• the modification of endpoints in v2.0 compared to v1.0 for 20 tRNAs

• changes in the claimed anticodon for 6 tRNAs

• the addition of 60 new human tRNAs in v2.0

Many new/modified human tRNAs have secondary structures that deviate substantially from the tRNA cloverleaf. We inspected manually the secondary structures of those human tRNA entries that are new to or have been corrected in v2.0 and found that

• at least 9 of the 41 elevated pseudo-tRNAs,

• at least 15 of the 20 entries whose endpoints changed in v2.0, and

• at least 18 of the 60 newly added tRNAs

(i.e. nearly 35% of the new/corrected entries) have abnormal secondary structures that deviate greatly from the tRNA cloverleaf.

Many v2.0 tRNA records are linked to incorrect legacy identifiers. 66% of the human tRNA records (405 of the 606) contained in gtRNAdb v2.0 have been linked to incorrect legacy identifiers. Specifically:

• legacy identifiers were given to the 60 new tRNAs of v2.0 even though they did not exist in v1.0

• 331 tRNAs whose coordinates did not change between v1.0 and v2.0 have been associated with a legacy identifier in v2.0 that does not match the original identifier in v1.0

• 14 entries that had their endpoints modified in v2.0 have been assigned legacy identifiers that do not match the entries’ original v1.0 identifier

Many v2.0 tRNA records contain data that are in conflict with their counterpart NCBI Gene and HGNC records. For 116 of the 606 human tRNAs, their gtRNAdb v2.0 records list different chromosome, strand, and endpoint information than the respective NCBI Gene record. These chromosomal location incompatibilities also extend to HGNC (HUGO Gene Nomenclature Committee) records that are linked to directly from within gtRNAdb records.

Competing interests: The above observations were compiled by Phillipe Loher, Venetia Pliatsika, Aristeidis G. Telonis, Yohei Kirino and Isidore Rigoutsos all of whom are with the Computational Medicine Center of Thomas Jefferson University and are actively involved in tRNA research or have published previously in this area. PL, VP, AGT, YK and IR declare no competing financial interests.

More information: a more detailed description on the above together with accompanying Figures and Tables can be found here.

On 2016 Jan 21, Isidore Rigoutsos commented:

Comments on the contents of release v2.0 of gtRNAdb

The gtRNAdb repository is a great resource for researchers studying transfer RNAs (tRNAs). With the increasing interest in the roles tRNAs and tRNA fragments play outside the confines of codon translation into amino acids, databases such as gtRNAdb are expected to provide invaluable reference information.

We focused on the Homo sapiens portion of gtRNAdb v2.0 and found the following:

Sweeping, undocumented changes in v2.0. In the above article, the authors state that they populated gtRNAdb v2.0 using a method that underwent major development but has not yet been peer-reviewed (the method is cited as “Chan et al., in preparation”). Specifically, for Homo sapiens the new method resulted in changes affecting 28% of the human tRNA records; the changes comprise

• the deletion of 74 tRNA records originally in v1.0

• the “elevation” of 41 pseudo-tRNAs of v1.0 to tRNAs in v2.0

• the modification of endpoints in v2.0 compared to v1.0 for 20 tRNAs

• changes in the claimed anticodon for 6 tRNAs

• the addition of 60 new human tRNAs in v2.0

Many new/modified human tRNAs have secondary structures that deviate substantially from the tRNA cloverleaf. We inspected manually the secondary structures of those human tRNA entries that are new to or have been corrected in v2.0 and found that

• at least 9 of the 41 elevated pseudo-tRNAs,

• at least 15 of the 20 entries whose endpoints changed in v2.0, and

• at least 18 of the 60 newly added tRNAs

(i.e. nearly 35% of the new/corrected entries) have abnormal secondary structures that deviate greatly from the tRNA cloverleaf.

Many v2.0 tRNA records are linked to incorrect legacy identifiers. 66% of the human tRNA records (405 of the 606) contained in gtRNAdb v2.0 have been linked to incorrect legacy identifiers. Specifically:

• legacy identifiers were given to the 60 new tRNAs of v2.0 even though they did not exist in v1.0

• 331 tRNAs whose coordinates did not change between v1.0 and v2.0 have been associated with a legacy identifier in v2.0 that does not match the original identifier in v1.0

• 14 entries that had their endpoints modified in v2.0 have been assigned legacy identifiers that do not match the entries’ original v1.0 identifier

Many v2.0 tRNA records contain data that are in conflict with their counterpart NCBI Gene and HGNC records. For 116 of the 606 human tRNAs, their gtRNAdb v2.0 records list different chromosome, strand, and endpoint information than the respective NCBI Gene record. These chromosomal location incompatibilities also extend to HGNC (HUGO Gene Nomenclature Committee) records that are linked to directly from within gtRNAdb records.

Competing interests: The above observations were compiled by Phillipe Loher, Venetia Pliatsika, Aristeidis G. Telonis, Yohei Kirino and Isidore Rigoutsos all of whom are with the Computational Medicine Center of Thomas Jefferson University and are actively involved in tRNA research or have published previously in this area. PL, VP, AGT, YK and IR declare no competing financial interests.

More information: a more detailed description on the above together with accompanying Figures and Tables can be found here.

On 2016 Feb 11, Todd Lowe commented:

We thank Isidore Rigoutsos and co-authors for the helpful, detailed comments. We have addressed all the issues in their commentary.

In brief:

1) The legacy names have been fixed and updated.

2) For now, we have put the pseudogene designations back into the database — this will likely change in the future as we develop a more objective, less arbitrary cutoff for what is called a tRNA pseudogene. We will post release notes when we do make these types of changes in the future, and welcome feedback from users.

3) The name mismatches between the GtRNAdb, NCBI, and HGNC for human tRNAs is due to an incomplete update of HGNC records requested previously, but we are coordinating with them to get these updated quickly.

4) We have put a disclaimer on the front page of the GtRNAdb letting users know that this database is a reflection of what we believe is the state of the art in tRNA detection and annotation. Our own understanding of tRNAs has been greatly enhanced by the use of the new isotype-specific models and increased sensitivity from tRNAscan-SE 2.0 (manuscript in preparation), so the criteria that we previously applied are, in some cases, clearly outdated and can be improved. A number of manuscripts are in preparation detailing these new insights based on our improved tRNA detection methods.

5) If users prefer a static, historical view of tRNAscan-SE gene calls, the prior GtRNAdb will still be available for reference at http://gtrnadb2009.ucsc.edu/ for the foreseeable future.

6) Some endpoints have indeed changed slightly (generally by 3-10 nucleotides total, for just 15 of 600+ total genes) because low-scoring tRNAs are aligned & scored slightly differently by Infernal 1.1 and the new tRNA covariance models, compared to the older software. We have over 60 new isotype-specific covariance models that we are still improving and refining, so small adjustments are expected over the next few months.

7) Of the "new" tRNAs in the GtRNAdb from the human genome, the vast majority are either very low scoring (i.e., they were just below the 20.0 bit covariance model score reporting cutoff used by the old version of tRNAscan-SE; with Infernal and new models, those scores may shift by 2-5 bits, bringing some above the reporting threshold, and some dropping below it). Because these differences are only affecting very low-scoring tRNAs, they appear to have little to no effect on the high-scoring tRNAs used in translation. Also, a fair number appear to be mitochondrial-derived tRNA genes, which tRNAscan-SE 2.0 is now able to detect with high sensitivity. These nuclear-encoded mitochondrial tRNA genes are now recognized to be commonly found in nuclear genomes, just as many mitochondrial proteins have migrated to the nuclear genome.

We regret any inconvenience these issues have caused users in the first few weeks since the database went live in December 2015. We encourage users to email us directly if they have questions or issues we can address. We are working hard to make this a useful, powerful resource to support the rapidly growing field of tRNA research.

Todd M. Lowe, Biomolecular Engineering, University of California, Santa Cruz