problems to pass from QIIME to MEGAN5

Question

[imindias]

Dear all,

First of all hello to everyone, this is my first post and I hope you can help me.

I'm an amateur in bioinformatics and I'm working in microbiota. We use 454 technology by Roche. To analyze diversities, I use qiime, it was difficult for me but finally I got it! but now I want to continue to the next step, the functionality.

I used to work with the tools that WebMGA gave me (fraggenescan, COG and KEGG databases) but I knew MEGAN and it seems very easy and with many opportunities for somebody like me that is learning now.

Well, please, don't think I am a completely idiot, but I am completely lost. I have used my out_table.biom from qiime into Megan and it's good, but I don't get new results, as I can't access to the COG and KEGG analysis.

So I tried to BLASTX my files, but it's impossible to use it (in fact, I am not sure if the files I am introducing are the correct), and I tried diamond, but I have few free space in my computer, and the analysis is killed in the middle.

I am completely desperate.

Please, can anybody indicate me the steps that I need to follow?

Thanks in advance!

[5heikki]

Do you have wgs or amplicon data? If latter, is it 16S?

[imindias]

Yes, I have 16S amplicons

[imindias]

Many thanks!

I'll try it and I let you know.

Thanks!!!

[imindias]

Hi again,

I tried Picrust and the galaxy version, and I am stuck again.

I have my file: predict_metagenome.picrustp, that acts as a text file. Is this a .biom file? I'm not sure how to proceed now.

Thanks!

[5heikki]

It's either a tsv or a biom file depending what output you specified. If you specified tsv (Legacy QIIME format (tab-delimited)), then one way forward is to edit out the first line and run the result through HUMAnN. Commands such as head and tail go a long way finding out what kind of files you're dealing with. IMO one of the core aspects of bioinformatics is understanding the nature of the data formats you're dealing with..

[imindias]

Dear Josh,

thank you for your so nice words. Sometimes, I feel as a complete idiot.

Well, I'm working with human gut microbiota, and I'm using the 454 technology by Roche.

I would like to use MEGAN, because I saw that permits a good visualization of, for example, KEGG pathways. But I'm open to another suggestion, taking into account my limited knowledge.

I started to sue WebMGA, because it's very simple and you don't need a good computer, but you only have some results, visualization is not possible and it's hard to analyze the data.

Thanks in advance

Isabel

[Josh Herr]

First of all, you're not an idiot -- you got a biom file out of QIIME, so you're certainly not an idiot by any means!

So, as 5heikki mentioned, PICRUSt would be a good way to address putative function for the human 16S amplicons. If you're having compute memory issues (join the club!), you can always try the Huttenhower Galaxy server for PICRUSt which also has some other tools for visualization. Give those a test drive and if you have some questions, feel free to post again or send a direct message. Best of luck!

Answer 1

[5heikki]

I would suggest closed-reference OTU-picking against Greengenes 13_8 (13_5) in QIIME and then load up the biom file into PICRUST for metagenome KO predictions. The galaxy version is super simple to use..

Answer 2

[Josh Herr]

Don't worry about being an amateur, we're here to help you! Everyone starts out at the beginning.

So, you have 16S data -- which only tells you about taxonomy in your system. I'm not certain what you want to show exactly, but you would like to show some functional information about your samples. As 5heikki mentioned, a good way to do this in by inferring function using your 16S data from the corresponding genomes via PICRUSt. I should warn that PICRUSt works well for human 16S sequencing where we have good representation of genomes to infer function from, but it doesn't work so well in environments (such as soil) where there are a lot of contributors not in the genome databases. You didn't tell us the type of data you have, so it's hard to recommend what to use.

I'm wondering if you'd like to use MEGAN to visualize your data -- MEGAN is a great BLAST visualizer, but there are other ways of looking at your biom file. I would highly recommend the R package phyloseq if you're looking to show your data structure. If you tell us a little about your data I could suggest more options -- there are many many to choose from.