Hi Vania,

If I understand your question correctly, your major issue is with your third library where you wish to segregate the reads of your focal mammal species from reads of other exogenous DNA present in the fecal DNA extract.

As you already mentioned you have a reference genome of your mammal species,

I suggest you map your raw data to the reference genome (I use BWA mem, but there are many other aligners available). Once the alignment is done, the reads which map to our reference are the ones of your mammal species and you can ignore all the reads which do not map properly.

I hope this might be of some help.

Thanks

Abhinav Tyagi : My assumption is the faecal reads which do not map to the mammal are the reads which need to be retained. The experimental objective is to use these retained reads as some kind of marker in order to determine which fish species (or species group) was consumed as prey by the mammal. Vania Carolina Fonseca da Silva has not detailed how she intends to use these retained reads.

I have a follow up comment proposing the approach I would take, and would be interested in comments on this approach.

Vania Carolina Fonseca da Silva You may want to consider a pangenomic approach whereby you use K-mer count sets with calculable probabilities of a faecal K-mer set originating from a fish species or group of species.

Because you have sampled enrichment you only have a probability of sampling any given SSRs in each species, and it is difficult to determine that probability for each fish species if the genome is unknown.

My suggestion is to take each set of fish reads and do a K-mer analysis, maximising the K-mer size until you obtain a distribution such that – guestimate – 10% of K-mers have coverage of between 10 and 100 copies

The guestimate of 10 is because you are expecting some coverage at the sampled SSRs and less than 10 could be indicative of sequencing errors. My guestimate of 100 is because you are looking to discard the SSR repeat polymerics and are looking for K-mers likely to be unique both intra and inter fish species. Hopefully for each fish species you will end up with many thousands of identified K-mers accepted within the copy number constraints and of maximal size up around 50% of your read lengths. You can’t use K-mer sizes up around read lengths simply because most K-mers will be discoverable through partial read overlaps, not full length overlaps.

Repeat iteratively over all fish species, adjusting the K-mer size until some minimum number of K-mers per species is reached, and build a matrix containing columns: fish species, K-mer sequence, count of that K-mer present in faecal sample reads – initialise count to 0.

With your faecal sample reads use a sliding window of the K-mer size you built the final accepted matrix with, and slide this window along each read.

For each K-mer from the sliding window check if that K-mer is present in the mammal genome – if so then discard and try next window.

If K-mer not in mammal genome then search matrix for a matching K-mer sequence and for every fish which matches then increment that fishes faecal K-mer match count.

When all K-mers in the faecal sample reads have been processed you can process the counts in the matrix and assign probabilities of the mammals individual fish species (or groups of fish) consumption.

I suggest that you discard counts of less than some threshold – say 10 - as these could represent sequencing errors.

The foregoing approach will need much optimisation to become practical, but could be applied even when the prey consumers genome has yet to be assembled providing you have reads from the consumer species. K-mers from the consumer reads are added to the matrix and so if count more than some threshold then discard that K-mer from further analysis!

I am interested in your research and may be able to provide some bioinformatics expertise ( https://github.com/kit4b ), do you have further details of your experiment you can share?