I would recommend the mothur approach. There are a few ways to create OTUs but I generally favour the method mothur uses unless the computational time/burden is going to become prohibitive. With 60 sequences that's not going to be a problem.

Also...you seem to be confusing classification with OTU assignment. They're both methods of clustering sequences but the underlying methodologies are very different. There's nothing wrong with doing both OTU assignment and classification on your data (it's a good idea to do both, to be honest), but just a general comment as they really are independent of each other. 

Article Assessing and Improving Methods Used in Operational Taxonomi...

I would recommend the mothur approach. There are a few ways to create OTUs but I generally favour the method mothur uses unless the computational time/burden is going to become prohibitive. With 60 sequences that's not going to be a problem.

Also...you seem to be confusing classification with OTU assignment. They're both methods of clustering sequences but the underlying methodologies are very different. There's nothing wrong with doing both OTU assignment and classification on your data (it's a good idea to do both, to be honest), but just a general comment as they really are independent of each other. 

Article Assessing and Improving Methods Used in Operational Taxonomi...

AS David told, Mothur will be the most reliable approach for OTU classification,.

I am also using mothur for OTU clustering (alignment to the SILVA SSURef taxonomy with the RDP naїve Bayesian rRNA classifier (Pruesse et al., 2007; Wang et al., 2007).

Thank every one for your advice. However, I have small problem that around  50% of sequencing results of the clone inserted in reverse orientation, so I am not sure that I have to remove or modify these sequence before running Mothur, if these sequences could be modified, could you please give suggestion to modify them. I am also the beginner in using Mothur and I have many problem. In Mothur wiki, they have only 2 examples that show how to use Mothur for MiSeq SOP and 454 SOP, while my sequences are Sanger sequences. So I wonder If I have to run trim.seqs command before clustering sequences into OTUs? I see to run this command I have to have obligos file and barcode, but I don't have these information.

Hi Pham, you don't have to remove any of the sequences before running mothur, but you can assemble forward and reverse sequences if you sequenced in both directions, did you? You can use http://www.codoncode.com/aligner/ (one month free trial). Create a fasta file with the sequence assemblies. Check with summary.seqs(fasta=xxx.fasta) if the file provided can be read by mothur.

For trim.seqs you first have to create an xxx.oligos file. You don't have barcodes, so just list primers used for sequencing in this file, for example:

forward CCGTCAATTCMTTTRAGT

reverse  CCGTCAATTCMTTTRAGT

Save the file (xxx.oligos). 

Now you can run the trim command. 

Then, you can allign them to the SILVA SSURef taxonomy with the RDP naїve Bayesian rRNA classifier and cluster into OTUs.

Thank you very much,  Evelyne Mann. You mean that I should separate my sequences into two files, one contains sequences inserted in reverse orientation, the other contains sequences inserted in correct orientation.

For the oligos file, do I need to create two file. For example:

One is:

forward CCGTCAATTCMTTTRAGT

reverse  CCGTCAATTCMTTTRAGT

The other is:

forward   CCGTCAATTCMTTTRAGT

reverse CCGTCAATTCMTTTRAGT

Hi Pham, 

no, it's the opposite around, as a first step you should put together forward and reverse sequences if you have not done this yet. As mentioned, you can use http://www.codoncode.com/aligner/ (one month free trial) for doing this. But there are many other tools, here are some more suggestions: 

http://www.researchgate.net/post/How_to_generate_consensus_DNA_sequence_contig_from_forward_and_reverse_sequence_Which_software_will_I_use 

Once you have your sequence assemblies,  you save them as a fasta file. 

Maybe you already have your assembled sequences, then you go on with creating a oligos file. 

Hope this helps...

 I remember that my clones were sequenced using only forward primer

M13F(-20) 5´-GTAAAACGACGGCCAGT-3´ 

 I already have assembled sequences, I will produce oligos file and run trim.seqs. My background is chemistry and I am not good in biology, so thank you very much.

Hi Evelyne Mann,

I have tried this method and Mothur said that: " your oligos file does not contain any group names. Mothur will not create the group file". Do you think it is still ok and I can continue do next steps.

The group file is just for when you want to allocate particular sequences to different sample groups. You don't need to have one, and if all your sequences came from a single clone library then there's no point in having one. For small numbers of sequences they're very easy to create manually if you decide you need one later on.

I lost track of this thread over the holidays, but I'm not sure you need to do the trim command at all. If you've already assembled your forward and reverse reads, then for OTU building you'll only need the following commands:

align.seqs()

screen.seqs()

filter.seqs()

dist.seqs()

cluster()

get.oturep()

If you're reading the 454 SOP, these are steps 4, 5 and 8.1. All the other steps at the start are to do with automatic error correction (denoising for 454 data, and quality filtering for MiSeq data).

Hi David W Waite,

I already tried these steps. However, I have some problems. Could you please help me?

When I run trim.seqs command for my sequence assemblies with oligos file, the trim.fasta file contains nothing.

Hence, I run align.seqs without using trim.seqs command. Do I need to use fasta format when run filter.seqs? when I run dist.seqs, Mothur said that:" It took 0 to calculate the distances of 52 sequences", so I think It might be one problem.

Hm, the trim issue could be a few things. First, did you do any manual quality filtering of the sequences before you started working in mothur? If so, you might have already clipped out the primer sequences (or part of it) which will mean mothur can't identify them in your sequence. Also, are you using the pdiffs parameter at all?

The trim.seqs command isn't essential, all it does is remove primer and barcode sequence and optionally performs quality filtering. On clone library data I've always done this manually. If you're happy with the quality of your sequences you can move straight into alignment.

You will need the fasta format for filter.seqs. The output from align.seqs will be a fasta file with your sequences stretched over the full 50,000 base position alignment space, and filter.seqs command removes the unnecessary spaces and removes hyper-variable positions. I couldn't tell you how long the dist.seqs command will take but I wouldn't be surprised if it was only a few seconds as mothur is optimised to work with very large data sets, so I wouldn't read much into the time taken as long as you can see that there is data in the output file (and it's not blank).

Hi David W Waite,

Thank you for your help.

I have not yet done any manual quality filtering of the sequences before you started working in mothur. I have not yet use the pdiffs parameter. 

For other trying, I use directly the file containing sequences without run trim.seqs command. I already run align.seqs, but the output of align.seqs is not fasta file. It is align file. So before run filter.seqs command, I copy content of output file and paste to text file. Then I use clustalX2 program to change the file to fasta file. Next I run the filter.seqs command for this fasta file. As a result, Mothur said that:" It took 0 to calculate the distances of 52 sequences". 

O, I see. The .align file is written in fasta format, I'm not sure why mothur changes the extension name. If you can see that there is data in the fasta file while you move through clustalX2 then the distance calculations will be fine. I just had a look through some old 454 data of mine and it took mothur 20 seconds to calculate the distances of 3212 sequences so your time sounds pretty normal.

ah, so you means when Mothur said:" It took 0 to calculate the distances of 52 sequences". 0 in here means 0 seconds. Oh, thank you very much, now I understand :D

Hi David W Waite,

I have tried and got the result of calculating OTUs for my sequences. During running Mothur, I realized some problem like that:

- When I align my sequences with silva.bacteria.fasta file, Mothur said that:"Some of you sequences generated alignments that eliminated too many bases. If the reverse compliment proved to be better it was reported". Do you think it still ok and I go to next steps?

This is the output of summary.seqs command after I align my sequences 

                        Start       End     NBases     Ambigs    Polymer  NumSeqs

Minimum:        1044    37682      1084           0                  4       1

2.5%-tile:      1044      37684       1098           0                 5       2

25%-tile:       1044      40159       1122           0                 5       14

Median:         1044     40339       1143            0                 6       27

75%-tile:       6202      43116       1158            0                 7       40

97.5%-tile:     6434     43116      1188             0                 8       51

Maximum:     6458    43116       1197              0               10      52

Mean:         3421.52 41216.5   1141.94           0           6.21154

# of Seqs:      52

These sequences are not align at same position, do you think it is because I have not yet run trim.seqs?

- For filter.seqs command, if I directly using output of align.seqs command, there was no problem, but if I use clustalX2 to change the file to fasta format, Mothur said that: " the sequence are not same length, please correct''. So do you think I can use directly output of align.seqs command for filter.seqs command in next step?

- If I directly using output of align.seqs command, finally I got rep.names files (0.03.rep.names, 0.04.rep.names, 0.05.rep.names,....0.36.rep.names). I wonder if 0.03.rep.names file means OTUs is defined as groups of sequences more than 97% similar.

Could you please give me some suggestion?

Would you mind if I send you my sequences to check whether my result is correct and not? Sorry because I bother you so much. 

 I send you my sequences at attached file. These sequences have already removed primer (primer is used for sequencing).

Here are some information I think it is good for you.

- I use two following primer set for running PCR:

 27F 5’-AGAGTTTGATCMTGGCTCAG-3’

1492R 5’-TACGGYTACCTTGTTACGACTT-3’

- I align my sequences with silver.bacteria.fasta. I send you silver.bacteria.fasta file at attach file.

- For dist.seqs command I use average distance method.

- I send you the output of Get.oturep command at attach file.

Thank you very much for you kind.

Don't worry about the reverse complement message. That's fine - sometimes you'll find there are sequences that can't be aligned at all but you don't have any of those.

The alignment is fine over your sequences, but there's always a little bit of difference at the ends of the sequences where you have a few extra bases in some sequences compared to others. That's just a fact of the way Sanger sequencing and the quality control process works. It's why you filter the sequences - to remove positions that aren't common to all sequence in your data set.

You're right about the 0.03 being the equivalent of 97% (3% difference is the same as 97% similarity).

I've just run your data through mothur and I get the same result as you. At 97% sequence similarity the sequences 9-7 and 9-38 form a single OTU and all the other sequences are separate.

Thank you very much for your kind, David W Waite. I will continuously waste time to find out the solution for these problems if I don't have your help.  

Hi David W Waite,

I wonder one more thing. In order to calculate relative abundance, do I will use RDP classifier tool (Wang et al., 2007) to classify OTUs first and then I calculate relative abundance for class level or family level based on the result after using RDP classifier tool?

I also wonder what is difference between the following command for alignment the sequences with silva.bacteria. fasta file:

1.pcr.seqs(fasta=silva.bacteria.fasta, start=11894, end=25319,

keepdots=F)

align.seqs(fasta=....fasta, reference=silva.bacteria.pcr.fasta)

2. align.seqs(fasta=.......fasta, reference=......fasta)

3.align.seqs(candidate=.....fasta, template=......fasta)

thank you very much.

Hi David W Waite, 

I am using Mothur 1.38.0 for NGS sequenced data analysis and i have done all the processes except dist.seqs. This command results in error showing, "your sequences are not of same length" and process aborted. How can i overcome this issue?