I've had good results using velvet/metavelvet for de novo assembly. It can take a bit of playing around to get the best parameters, but since it's command-line based it's quite easy to re-run it many times with different settings. There are some easier to use software suites out there, like CLC, but in that case the licence is quite expensive so if you only have a single data set to analyse then it's probably not a realistic option.

Personally I prefer to do the assembly locally, but that's just my choice, and it is heavily influenced by whether or not you have the computational capacity to do this kind of analysis. If you're using a standard issue desktop PC with 8-16 GB of RAM you'll probably have to go with the online option. https://usegalaxy.org/ can do assembly (actually, it uses velvet too) although I haven't used it myself, so I don't know what sort of registration is required.

https://docs.google.com/document/d/1xs-TI5MejQARqo0pcocGlymsXldwJbJII890gnmjI0o/pub

Article Velvet: Algorithms for De Novo Short Read Assembly Using De ...

Article Namiki T, Hachiya T, Tanaka H, Sakakibara Y.. MetaVelvet: an...

I've had good results using velvet/metavelvet for de novo assembly. It can take a bit of playing around to get the best parameters, but since it's command-line based it's quite easy to re-run it many times with different settings. There are some easier to use software suites out there, like CLC, but in that case the licence is quite expensive so if you only have a single data set to analyse then it's probably not a realistic option.

Personally I prefer to do the assembly locally, but that's just my choice, and it is heavily influenced by whether or not you have the computational capacity to do this kind of analysis. If you're using a standard issue desktop PC with 8-16 GB of RAM you'll probably have to go with the online option. https://usegalaxy.org/ can do assembly (actually, it uses velvet too) although I haven't used it myself, so I don't know what sort of registration is required.

https://docs.google.com/document/d/1xs-TI5MejQARqo0pcocGlymsXldwJbJII890gnmjI0o/pub

Article Velvet: Algorithms for De Novo Short Read Assembly Using De ...

Article Namiki T, Hachiya T, Tanaka H, Sakakibara Y.. MetaVelvet: an...

Hi Alejandra, I would recommend to install assembler on your machine. Generally, meta-velvet is suited for assembling metagenomics reads. Although, a lot of caution needs to be accounted while doing the phylogenetic binning of reads or chimeric contigs developed while assembly is done. Post assembly, you may need to separate the contigs based on blast results. You may also check the A5 assembler or RAY assembler for this purpose. Thanks!

http://www.plosone.org/article/info%3Adoi%2F10.1371%2Fjournal.pone.0042304

Hi Alejandra, I would recommend the IDBA_UD (http://i.cs.hku.hk/~alse/hkubrg/projects/idba_ud/) ,a iterative De Bruijn Graph De Novo Assembler for Short Reads. It iterates from small kmer to a large kmer.

Hello David, Aamir and Guojun!

Thank you very very much for your answers, time, experience and recomendations, it is good to know about all the available tools we can use for metagenome assembly.

Greetings! :D

Take a look at the MetAMOS/iMetAMOS pipelines for metagenomic assembly. (http://metamos.readthedocs.org/) In addition to handling a number of steps before and after metagenome assembly, iMetAMOS lets you easily test most major metagenome assemblers (including all those listed in previous answers) in a single automated step. The authors also distribute a preconfigured virtual machine so you do not need to configure the dozens of software packages necessary for this sort of automated multi-assembler testing.

Thanks for your recommendation Jonathan! :) 

Hi Alejandra,

I would also recommend metavelvet and Ray. Ray can take advantage of Open MPI and truly parallelize the assembly across nodes, so you don't need to have a large memory node to do the job (provided that you have enough of them). It is specially useful if you have a large dataset and access to a computer cluster. For local assembly you could also try SOAPdenovo.  

http://soap.genomics.org.cn/soapdenovo.html

Thank you very much Pablo! :) 

I think CLCbio and Ray had better efficiency in memory usage, and IDBA-UD and Ray are faster because of the parallel. To maximum the genome recovery from metagenome,  my way is to individually carry out de novo assembly by IDBA-UD, CLCbio and MetaVelvet and then merge the three results by CLCbio or Soapdenovo with mapping algorithm.

Thank you Yu! :) 

You may also check out Genovo, a de novo metagenome assembler. Bambus 2 is a metagenome scaffolder which might be of help to you too.

http://online.liebertpub.com/doi/abs/10.1089/cmb.2010.0244

http://amos.sourceforge.net/wiki/index.php/Bambus2

Thanks for the links Dr. Xinkun Wang :)