lis4matilda
commented
6 years ago I think we should wait with that implementation and try to finish a database based workflow first. After that we can implement an assembly approach to create a database to use in this workflow.
16 apr. 2018 kl. 09:59 skrev Fredrik Boulund notifications@github.com:
We need to figure out how we want to implement metagenome assembly. The discussion needs to be divided into several parts:
What assembler(s) we want to use How to make the Snakemake technical implementation it so that any assembler can be used to produced output files for steps downstream of the assembly What steps to add downstream of assembly
- What assemblers to use
Several options are available to us. According to CAMI, assemblers like MEGAHIT performed fairly well. It seems SPADes was not included in the evaluation, but it seems like a popular assembler nonetheless. I suggest we start with adding these, unless you or anyone else has other preferences
- How to implement it
Perhaps it is easiest if we let each assembler output into its own folder, but move the assembled contigs into a folder specifically for assembled contigs:
output_dir/spades ├── sample1 │ ├── contigs.fa │ ├── file1.txt │ └── file2.fq └── sample2 ├── contigs.fa ├── file1.txt └── file2.fq output_dir/megahit ├── sample1 │ ├── contiga.fa │ ├── file1.txt │ └── file2.fq └── sample2 ├── contiga.fa ├── file1.txt └── file2.fq output_dir/assembled_contigs ├── sample1.contigs.megahit.fa ├── sample1.contigs.spades.fa ├── sample2.contigs.megahit.fa └── sample2.contigs.spades.fa A structure like this should make it fairly straightforward for us to design downstream steps that just process the contig files found in output_dir/assembled_contigs/, making it flexible enough so that users can pick whatever assembler they want, and also make comparisons between assemblers if that is interesting. It has the potential drawback of having to run all downstream steps that depend on assembly for the output from all assemblers used, but I guess most users will just pick one assembler and stick to that.
- What steps to add downstream of assembly
Typical steps we could put here are binning steps, to bin the assembled contigs into "metagenome-assembled genomes" (MAGS)/"metagenomic species" (MGS), which can then be further analyzed. Two popular alternatives here could be MaxBin and CheckM. Potentially, the CheckM-related genome binner GroopM could be used if samples are taken from "at least 3 timepoints" (according to their documentation), which might make it slightly difficult to implement in StaG-mwc as a general tool, because the assumptions that GroopM makes probably fail for most of our study designs.
I know there was some talk about co-abundance clustering to produce co-abundance gene groups (CAGs), exactly how were you thinking about that @lis4matilda?
— You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub, or mute the thread.
boulund
We need to figure out how we want to implement metagenome assembly. The discussion needs to be divided into several parts:
1. What assemblers to use
Several options are available to us. According to CAMI, MEGAHIT performed fairly well. It appears SPADes was not included in the evaluation, but it seems like a popular assembler nonetheless. I suggest we start with adding these, unless someone has other preferences.
2. How to implement it
Perhaps it is easiest if we let each assembler output into its own folder, but move the assembled contigs into a folder specifically for assembled contigs, for example (the actual outputs from the two assemblers in the example look different in real life):
A structure like this should make it fairly straightforward for us to design downstream steps that just process the contig files found in
output_dir/assembled_contigs/, making it flexible enough so that users can pick whatever assembler they want, and also make comparisons between assemblers if that is interesting. It has the potential drawback of having to run all downstream steps that depend on assembly for the output from all assemblers used, but I guess most users will just pick one assembler and stick to that.3. What steps to add downstream of assembly
Typical steps we could put here are binning steps, to bin the assembled contigs into "metagenome-assembled genomes" (MAGS)/"metagenomic species" (MGS), which can then be further analyzed. Two popular alternatives here could be MaxBin or CONCOCT, with a follow-on checkup of bins using CheckM. Potentially, the CheckM-related genome binner GroopM could be used if samples are taken from "at least 3 timepoints" (according to their documentation), which might make it slightly difficult to implement in StaG-mwc as a general tool, because the assumptions that GroopM makes probably fail for most of our study designs.
I know there was some talk about co-abundance clustering to produce co-abundance gene groups (CAGs), exactly how were you thinking about that @lis4matilda?