If you were a researcher interested in investigating a particular group of prokaryotes (e.g. Cyanobacteria) from an environmental sequencing project, which of the following approaches to taxonomy (using tools CLARK, Kraken, Centrifuge, Kaiju, etc. which utilize k-mer information) would you use?
I am working independently with a colleague who used Approach 2 and only got one species hit while I worked using Approach 1 and got a very diverse result (several (>20) species are present according to my result).
My colleague is firm with the results of their work. I, however, am not entirely convinced that there is only one species in the sample since we are doing classical isolation and we were able to get several unialgal isolates already. I would like to know which one of us is correct.
Would constraining the database from the start put a bias to the classification? Why do we have such diverging results?
Before commenting on the differences in results, I would say there is something wrong in either of your database creation. If one is using entire database and other is using specific part of the same database, both must get the same/similar results. The main difference here would primarily be the size of the database, as complete genbank database has everything other than cyanobacteria too.
Now coming to the results, the information you provided is not sufficient to say who is correct. I don't see much problem in the tool usage as long as you both have used same bioinformatics pipeline.
Approach 2 looks logical, as if you want to see cyanobacteria, why not just use cyanobacteria as specific custom database.
Approach 1 should have resulted in similar results if the cyanobacterial custom database was correct. Now such wider differences in this approach might question the threshold and percentage similarity used for the clustering.
You both need to dig deeper and see the differences in your analysis, then only there can be more specific questions and that might be answered further.
Thanks for your input Abhijeet Singh
I am actually surprised/weirded out by our results since you are right that they should be identical given the same GenBank database used. The other one should just be faster since a subset database is used. I am currently running both using exactly the same tools (contigs from metaSPAdes, and taxonomy using CLARK) and using databases indexed using CLARK as well.
- Badges
- Science method
- Similar topics
- Biological Science
- Genetics
- Genetic Analysis
- genetic techniques
- Sequencing