Shotgun analysis: Where and how to start?

Question

I am a complete rookie in shotgun analysis, having only performed marker gene analysis before. I have a pipeline that I want to try but I can't wrap my head around how it all works.

Please correct me if I'm wrong:

1. I start with a paired-end Illumina short reads fastq files (.R1,.R2)

2. I will use FASTQC to trim according to quality. I assume that since data has been demultiplexed, the primers and and adaptors are removed as well.

? First question, a lot of people advise to merge all the samples together and then do the co-assembly since apparently it helps with the assembly. Would this be necessary if I also have nanopore reads and only 9 samples? If if I did this, how do I then separate the samples from each other?

3. Because I also have nanopore, I will use a hybrid assembler to get contigs. The output will be a file with contigs.

4. I then plan to use a binning tool (Metawrap) to create bins.

? I believe the output of these will be a file containing bins per sample. How do I estimate the abundance of each bin in the sample? And how do I factor in the depth of sequencing to normalise them? This is my biggest problem, I can't wrap my head around it.

?Can some explain VERY plainly what is sequencing depth and how do I factor it into my analysis? I've seen some people using it for normalisation of the bin/contigs abundance but again, this is idea is not intuitive to me at all.

5. From then on, taxonomy and functional annotation is pretty straightforward.

#Update

So a little background on the project. I have samples sequenced from a bioreactor containing a small community of bacteria (no host removal required). The samples derive from 6 different reactors = 6 samples as well as one reactor will be samples three times (time series) so I have 9 samples in total. The shotgun Illumina will be combined with nanopore. The idea is to get the composition of the community (what kind of bacteria are there?), abundance (what is their relative quantity?) but also functions (what do they do?).

I'm sorry for a very generic question, I would really appreciate somebody answering very plainly without too much jargon. Hopefully it will help me and other people who are just starting out with the shotgun.

Thank you!

Answer 1

Question 1

For co-assembly vs individual assembly, it is sometimes advisable to do them separately, perhaps if you expect samples to be very different or if you have huge amounts of data. For <10 species I imagine that you will have quite high coverage, especially with nanopore.

Some metagenomic binners (e.g. MetaBAT2, a standard from a few years ago) can even incorporate information across samples, which might argue for maintaining at least some separation.

"How do I separate the samples from each other" I guess the point is that you don't, for assembly at least. You can then go in post hoc with your sequencing data from each sample and try to figure out with abundance analysis which of your genomes are in each sample.

Question 2

For abundance analysis/depth, there are strong statistical parallels in RNA-seq analysis, as Lior Pachter is fond of saying, if that is helpful. Any number of metagenomic shotgun abundance computations can be used to estimate abundance of genomes, including Pachter's metakallisto (linked therein). I found metakallisto fairly straightforward to use on contigs resulting from assembly.

Basically, you have some set of DNAs in your sample with some stoichiometry (relative abundance with respect to each other). The idea is to use DNA sequencing to estimate that stoichiometry (absolute abundance is much harder so most people don't bother).

Here is a broad view of the workflow:

1. You come up with some estimate of the DNAs present in the community. This can be a shotgun assembly, or some set of reference genomes, or whatever.

2. You count how many sequencing reads derive from each of those DNAs, usually via alignment or pseudoalignment.

3. YOu do some numerical corrections for systematic covariates such as DNA length. (These can look like RPKMs or TPMs from RNA-seq; metakallisto literally uses TPM calculations)

4. The resulting counts should be abundances, more or less.

You don't have to figure this stuff out yourself. There are various end to end pipelines that do this.

You will almost certainly need to compute abundances of contigs if you want to do metagenomic binning.