RDP reference files

Version 19

The publicly released version 19 of its training set in July 2023. You’ll notice that the names of the bacterial phyla and other levels have been significantly changed from prior versions. We have modified the files that they make available on SourceForge to be compatible with mothur. To maintain a consistent 6 taxonomic levels we have removed the various sub-classes, orders and families:

  • 16S rRNA reference (RDP): A collection of 23,853 bacterial, 788 archaeal, and 1 eukaryotic SSU rRNA gene sequences with an improved taxonomy compared to version 16.
  • 16S rRNA reference (PDS): The RDP reference with 119 mitochondrial 16S rRNA gene sequences added as members of the Rickettsiales and four 18S rRNA gene sequences added as members of the Eukarya.

You should be aware of several things when using the RDP training set. First, the taxonomies only go to the genus level; therefore, you will only be able to classify your sequences to the genus level. You can modify the training set to include species-level names and may be successful in classifying to the species level. Second, many of these sequences are very poor in quality. Low quality reads have a large number of ambiguous base calls or are very short. Here is the output from running summary.seqs on trainset19_072023.rdp.fasta:

           Start	End	NBases	Ambigs	Polymer	NumSeqs
Minimum:	1	425	425	0	4	1
2.5%-tile:	1	1310	1310	0	5	617
25%-tile:	1	1427	1427	0	5	6161
Median: 	1	1466	1466	0	5	12322
75%-tile:	1	1497	1497	0	6	18482
97.5%-tile:	1	1549	1549	10	7	24026
Maximum:	1	1968	1968	130	100	24642
Mean:	        1	1452	1452	1	5

Version 18

The publicly released version 18 of its training set in July 2020 (a version 17 was not released). We have modified the files that they make available on SourceForge to be compatible with mothur. To maintain a consistent 6 taxonomic levels we have removed the various sub-classes, orders and families:

  • 16S rRNA reference (RDP): A collection of 20,712 bacterial and 601 archaeal 16S rRNA gene sequences with an improved taxonomy compared to version 16.
  • 16S rRNA reference (PDS): The RDP reference with 119 mitochondrial 16S rRNA gene sequences added as members of the Rickettsiales and four 18S rRNA gene sequences added as members of the Eukarya.

You should be aware of several things when using the RDP training set. First, the taxonomies only go to the genus level; therefore, you will only be able to classify your sequences to the genus level. You can modify the training set to include species-level names and may be successful in classifying to the species level. Second, many of these sequences are very poor in quality. Low quality reads have a large number of ambiguous base calls or are very short. Here is the output from running summary.seqs on trainset18_062020.rdp.fasta:

           Start	End	NBases	Ambigs	Polymer	NumSeqs
 Minimum:	1	455	455	0	4	1
 2.5%-tile:	1	1315	1315	0	5	530
 25%-tile:	1	1426	1426	0	5	5299
 Median: 	1	1464	1464	0	5	10598
 75%-tile:	1	1493	1493	0	6	15897
 97.5%-tile:	1	1547	1547	12	7	20666
 Maximum:	1	1968	1968	130	100	21195
 Mean:	1	1450	1450	1	5

Version 16

The publicly released version 16 of its training set in February 2016 (we dropped the ball on version 15). We have modified the files that they make available on SourceForge to be compatible with mothur. To maintain a consistent 6 taxonomic levels we have removed the various sub-classes, orders and families:

  • 16S rRNA reference (RDP): A collection of 12,681 bacterial and 531 archaeal 16S rRNA gene sequences with an improved taxonomy compared to version 14.
  • 16S rRNA reference (PDS): The RDP reference with 119 mitochondrial 16S rRNA gene sequences added as members of the Rickettsiales and four 18S rRNA gene sequences added as members of the Eukarya.

You should be aware of several things when using the RDP training set. First, the taxonomies only go to the genus level; therefore, you will only be able to classify your sequences to the genus level. You can modify the training set to include species-level names and may be successful in classifying to the species level. Second, many of these sequences are very poor in quality. Low quality reads have a large number of ambiguous base calls or are very short. Here is the output from running summary.seqs on trainset16_022016.rdp.fasta:

       Start   End NBases  Ambigs  Polymer NumSeqs
 Minimum:  1   320 320 0   4   1
 2.5%-tile:    1   1267    1267    0   5   331
 25%-tile: 1   1424    1424    0   5   3304
 Median:   1   1465    1465    0   5   6607
 75%-tile: 1   1494    1494    0   6   9910
 97.5%-tile:   1   1549    1549    10  7   12882
 Maximum:  1   2210    2210    130 68  13212
 Mean:     1   1445.52 1445.52 1.0632  5.56282

Version 14

The publicly released version 14 of its training set in May 2015 (no idea what happened to releases 11 through 13). We have modified the files that they make available on SourceForge to be compatible with mothur. To maintain a consistent 6 taxonomic levels we have removed the various sub-classes, orders and families:

  • 16S rRNA reference (RDP) : A collection of 10,244 bacterial and 435 archaeal 16S rRNA gene sequences with an improved taxonomy compared to version 10.
  • 16S rRNA reference (PDS): The RDP reference with 119 mitochondrial 16S rRNA gene sequences added as members of the Rickettsiales and four 18S rRNA gene sequences added as members of the Eukarya.

You should be aware of several things when using the RDP training set. First, the taxonomies only go to the genus level; therefore, you will only be able to classify your sequences to the genus level. You can modify the training set to include species-level names and may be successful in classifying to the species level. Second, many of these sequences are very poor in quality. Low quality reads have a large number of ambiguous base calls or are very short. Here is the output from running summary.seqs on trainset14_052015.rdp.fasta:

             Start End NBases  Ambigs  Polymer NumSeqs
Minimum:   1   320 320 0   4   1
2.5%-tile: 1   1047    1047    0   5   267
25%-tile:  1   1426    1426    0   5   2670
Median:    1   1468    1468    0   5   5340
75%-tile:  1   1495    1495    1   6   8010
97.5%-tile:    1   1550    1550    12  7   10413
Maximum:   1   2210    2210    91  68  10679
Mean:      1   1443.09 1443.09 1.25714 5.58404

Version 10

The publicly released version 10 of its training set in September 2014. We have modified the files that they make available on sourceforge to be compatible with mothur. To maintain a consistent 6 taxonomic levels we have removed the various sub-classes, orders and families:

  • 16S rRNA reference (RDP): A collection of 10,240 bacterial and 410 archaeal 16S rRNA gene sequences with an improved taxonomy compared to version 9.
  • 16S rRNA reference (PDS): The RDP reference with 119 mitochondrial 16S rRNA gene sequences added as members of the Rickettsiales and four 18S rRNA gene sequences added as members of the Eukarya.

You should be aware of several things when using the RDP training set. First, the taxonomies only go to the genus level; therefore, you will only be able to classify your sequences to the genus level. You can modify the training set to include species-level names and may be successful in classifying to the species level. Second, many of these sequences are very poor in quality. Low quality reads have a large number of ambiguous base calls or are very short. Here is the output from running summary.seqs on trainset10_082014.rdp.fasta:

             Start End NBases  Ambigs  Polymer NumSeqs
Minimum:   1   320 320 0   4   1
25%-tile:  1   1426    1426    0   5   2663
Median:        1   1468    1468    0   5   5326
75%-tile:  1   1495    1495    1   6   7988
97.5%-tile:    1   1550    1550    12  7   10384
Maximum:   1   2210    2210    91  68  10650
Mean:      1   1443.04 1443.04 1.25822 5.5846  

Version 9

The RDP publicly released version 9 of its training set in March 2012. We have modified the files that they make available on sourceforge to be compatible with mothur. To maintain a consistent 6 taxonomic levels we have removed the various sub-classes, orders and families:

  • 16S rRNA reference (RDP): A collection of 9,665 bacterial and 384 archaeal 16S rRNA gene sequences with an improved taxonomy compared to version 7 (there was no v.8 as far as we are aware).
  • 16S rRNA reference (PDS): The RDP reference with 119 mitochondrial 16S rRNA gene sequences added as members of the Rickettsiales and four 18S rRNA gene sequences added as members of the Eukarya.

You should be aware of several things when using the RDP training set. First, the taxonomies only go to the genus level; therefore, you will only be able to classify your sequences to the genus level. You can modify the training set to include species-level names and may be successful in classifying to the species level. Second, many of these sequences are very poor in quality. Low quality reads have a large number of ambiguous base calls or are very short.

Version 7

The RDP released version 7 of its training set in November 2011. In separate files they provide the reference data for 16S (Bacteria and Archaea) and 18S (fungi) rRNA gene sequences and taxonomy. We have modified these files to be compatible with mothur. To maintain a consistent 6 taxonomic levels we have removed the various sub-classes, orders and families:

  • 16S rRNA reference (RDP): A collection of 9,662 bacterial and 384 archaeal 16S rRNA gene sequences with an improved taxonomy compared to version 6.
  • 16S rRNA reference (PDS): The RDP reference with three sequences reversed and 119 mitochondrial 16S rRNA gene sequences added as members of the Rickettsiales
  • 28s rrna reference (rdp): A collection of 8506 reference 28S rRNA gene sequences from the Fungi that were curated by the Kuske lab

You should be aware of several things when using the RDP training set. First, the taxonomies only go to the genus level; therefore, you will only be able to classify your sequences to the genus level. You can modify the training set to include species-level names and may be successful in classifying to the species level. Second, many of these sequences are very poor in quality. Low quality reads have a large number of ambiguous base calls or are very short. In the PDS version of the training set we have reversed three sequences that were in the wrong direction.

Version 6

The rdp training set (version 6, released 03/02/2010) consists of 8,422 sequences (8,127 bacterial and 295 archaeal) and is based on Bergey’s taxonomic outline. This training set is our modification of the files that they posted to sourceforge. Their archive provides a bunch of other files that mothur will calculate the first time you use the training set. Our archive consists of two files - a fasta-formatted sequence file and a mothur-compatible taxonomy file. The only subtle manipulation we made was to remove the sub-taxonomic levels (e.g. sub-order) and to plug in incertae_sedis levels when a step in the taxonomy was missing. Thus taxonomic level 6 corresponds to the level of genus and level 1 corresponds to the level of kingdom. I have also included a “pds” version of the same reference collection that includes additional Mitochondrial and Eukaryotic sequences.