Subset your taxa of interest and than proceed with your analysis if you dont want to consider the relative abundance of any other except the ciliates.

The molecular data bases on ciliates are far away from being complete or better, they are very incomplete.

If your director is only molecular-man, you have lost. I don agree that for today, molecular tools applied alone are the best way to study the ciliate communities. In the future it shall be much more important but still, we are just parching bases of our knowledge combining new and old methods.

If your director is a good ciliatologist, try to run any quantitative taxonomical method to compare with the results as OTUs. Be aware to evaluate OTUS like abundances because ciliates varied by orders of magnitude in their nucleic acid content (just now I am fighting with it in environmental samples). I am using Quantitative Protargol Stain by Montagnes and Lynn, applying some modifications (mainly heat) published by Skibbe.

And compare and evaluated and be "ethic" in interpretations.

Good luck

Miroslav Macek

P.S. The Kindom of Protista (Protozoa) is over. I am a republican.

Miroslav Macek

Hi Miroslav,

I am very glad you answered!

Currently I did the same: I have morpholgical data with QPS of the same samples as my molecular data. But I don´t know what to do now?

I hoped i would have like a inshore and offshore subset (like Santoferra et al 2016) and be able to compare the composition of these subsets between molecular and morphological data. But I don´t have this clear structure in my communities.

My prof had the idea of making a CCA with the environmental paramteres and see if both molecular and morphological data are comparable. But since I am once reducing 1300 (different otu´s) to 2 dimensions and once 100 (species from QPS) to 2 dimensions I can hardly compare it.

I thought about a Mantel test to take a look if these 2 matrizes are sharing same structure?

So what exactly do you do with the data of the OTUs and the QPS data?

I think that may sound like a stupid question, I know, I am feeling the same way, but can´t help it, I don´t know how I shall compare the data...

Do you perhaps have any further ideas what I could do with my data?

Thank you very much, Anita,

I think you are in the perfect hands (of your director).

On the other hand, I am not able to help you with the molecular part (my maximum of molecular methods were FISH and CARD-FISH methods applied in my lab to study ciliate vacuole content). Other publications were made with my prokaryote-colleagues (again, limnology & microscopical methods from my part). Now, the student of mine will look for molecular characterization of anoxic ciliate (potential) cyanobacterial-symbionts, however, this part of her work will be managed in the lab of my colleague. I have not got the lab equipped for such methods, I will die in front of the microscope...

Since 2014 I has been analysing protargols from ~100 lakes in Finland and now they will try compare RNA data with those of mine. Just today I am sending the enhanced data set including biomasses. However, until now I do not know anything about the molecular results.

As you can see, I am not against such methods; on the other hand, I prefer to share my data with my colleagues specialised on molecular biology, even from other countries (I am of Czech origin, working in Mexico already for a quarter of century).

If you are interesting to continue exchanging information, I will be glad.

Best wishes from Mexico

[email protected]

P.S. Who is your supervisor? Somebody from Sabine or Ulrike group?

@ Abhijeet Singh : Thank you for your answer! I started with rarefying my taxa of interest and came to a new problem-

may I go more in detail, please?

My taxa of interest are on the one hand ciliates on the other hand dinoflagellates.

Most of the analyses I perform only on the ciliates. If I look at the difference in average library size for the ciliates only I do have 17.5 x and for the dinoflagellates together with the ciliates I do have 7.3x.

So my questions:

Do I do rarefying for each taxa or for the both taxa as one otu table? If I want to compare the abundance of the two taxa among each other, I need to rarefy both at one time, right? But due to the huge differences in the average library size I am unsure about it...

If I rarefy only ciliates I "loose" 31% of ciliate otu´s, if I rarefy ciliates and dinoflagellates I "loose" 18% of the ciliate otu´s...

which bioinformatic pipeline are you using for your data analysis? And do you have any reference database if you want to assign the taxonomy ?

Additionally, please find the attachment regarding rarefying your data

@ Abhijeet Singh

Sorry that it took me so long to answer. Thank you very much for your time: I hope I can provide you some of the missing information.

The pipeline: Trimmomatic >PEAR>Cutadapt>vsearch>swarm> mothur/RDP Classifier with silva128ref and pr2_gb203v4.5

Thank you for posting the "Waste Not, Want Not..." paper. I decided to give rarefy a try, because of the following paper:

Weiss, Sophie, et al. "Normalization and microbial differential abundance strategies depend upon data characteristics." Microbiome 5.1 (2017): 27.

Since the ciliates have an average library size of 17,3x, I thought rarefy could be the right choice, since the paper suggests, that the method used should be choosed according to the data characteristics...so for me it sounded plausible, but I am only a beginner and glad, if you share your experience. What technique do you use?

For me the other suggested techniques Negative

Binomial method and zero-inflated Gaussian mixture from the packages EdgeR and phyloseq (as far as I understood) ask for treatments and biological replicates, which unfortunately we don´t have (expect for one sample, where we were only able to filter 100ml because the "organisms" were that dense)... That´s also why I didn´t use them. Can I use them without having biological replicates/ without having different treatments and do they perform well, too if this information is missing?

How can I translate that in R, do I need a vector that concludes that every sample is a different treatment?

* The pipeline: Trimmomatic >PEAR>Cutadapt>vsearch>swarm> mothur/RDP Classifier with silva128ref and pr2_gb203v4.5

> I am a bit confused to see that you used Trimmomatic and after that cutadapt, although they are doing almost same thing to the data. Moreover, you could have used --fastq_mergepairs in vsearch instead of using PEAR.

Your pipeline is confusing for me. If I would be you I would do as follows

fastqc > cutadapt > fastqc > vsearch for clustering > taxonomic_assignment either by mothur or qiime (*)> phyloseq

(* - why do you use two databases ? I recommend to use anyof them but only one use silva with silva classifier of RDP with RDP classifier.

If you are using mothur as you mentioned you can just use following for the taxonomic assignment with silva. https://www.mothur.org/wiki/Classify.seqs)

emothur > classify.seqs(fasta=YOUR_OTU_sequence.fasta, template=silva_referenc.fasta, taxonomy=silva.bacteria.silva.tax)

* I haven't worked with EdgeR so no comments. But, there is no limitation/requirement/boundation in phyloseq for using triplicates. It will just use the parameters what you define in your mapping file. If you don't have triplicate, it does not matter. You can still use phyloseq and visualize your analyzed data.

Abhijeet Singh thank you for your answer!

Sorry that I took so long to reply.

I didn´t do the bioinformatics ( and can´t do it because the deadline for my master thesis is that close)- so I had to ask for more information but only got that far:

The raw data was trimmed with Trimmomatic, default settings used for AdapterClipping. Slidingwindow was performed with length 4bp with a minimum average phred quality score of 8 for verifying the quality of the single reads and cutting at the 3’ end once the average quality within the window falls below the threshold. With PEAR the paired ends of the reads were merged with a minimum overlap of 10bp. With Cutadapt the adjustment of the reads was filtered according to the model: Forward primer followed by the read and ending with the reverse primer. As a threshold at least 75% of the primer must be found or a maximum of 25% of the bp may be mismatches. Subsequently the primers have been removed using Cutadapt. With VSearch the reads with an expected error bigger than 1 have been filtered. The error has been calculated due to the phred scores of the fastqs. Furthermore reads with ambi-bases or deviations in length bigger than +/- 50bp from the mean of all reads were discarded. Chimera were identified de novo with Vsearch using default values and removed. OTU clusters with distance 1 and boundary 2 were identified using Swarm and singletons were removed. For mother /RDP classifier the references silva128ref and pr2_gb203v4.5 have been used in parallel with a confidence interval of 90%.

So I cannot change the pipeline anymore but would be happy for comments to discuss in my thesis, since I have no experience using the different softwares. Also I think with 10 000 dinoflagellate OTUs and 1500 ciliate OTUs our clustering or chimera detection or... was not adapted perfectly to our data, it sounds really high compared to other studies.

I asked for more information but so far these are all I got.

What would you say to our approach of clustering swarm distance 1 and boundary 2?

Best wishes,

Anita