Hello everyone,
I have some questions in regards to using BLAST and sequencing data to give a species I.D. for fungal isolates.
We have created pure isolates of fungal colonies and then sequenced a gene region (shown in previous research to be a suitable DNA barcoding region for this genus) and ran this region (after sequencing both the forwards and reverse, creating a consensus sequence and trimming) through the BLAST database. While most isolates come back with mainly one species in the top search results, some have come back with various species in the top search results.
Therefore, how do researchers determine which species is the more appropriate I.D. using the BLAST database, or would you just say it is ambiguous based on the results?
For example, a general rule of thumb I was told in my undergraduate degree was you need an e-value below 10-4 and a percentage identity above 70%, however, if there are multiple species with similar values in the search results, how do you then decide which is most appropriate?
Is it purely based on the numbers, or would you look at the source of the result and preferably use isolate collections? Or, is it perhaps that newer species will have fewer entries in the database, so other species are likely to appear in the search results?
Thank you for your time,
Lauren
Hello, for understanding BLAST i suggest you watching this video, it may helpful for you for have a vision in short term
https://www.youtube.com/watch?v=bL-GnXJHdek
Dear Lauren,
In this case it is better to say that identifications are ambiguous. Having multiple hits with the same (low?) similarity means that your sequences are not very similar to anything that is so far available in the database.
One way around this would be to download the most similar sequences (or what you think is taxonomically closer to your focal taxa) and build a phylogenetic tree. Such a tree would give you a more precise insight into the taxonomic affinity of your taxa than just a simple percentage threshold.
It is possible that your species is not represented in NCBI database, therefore you obtain low % identity values, shared for different species. It depends on the taxonomic group and genetic marker, but % identity 95% can be misleading, there may be errors in deposited sequences. You should try to base your identification in a high %identity, not shared with different species and obtained in different studies (several entries in the database). Ideally, more than one genetic marker should be used with similar results.
Give also a look into: https://www.boldsystems.org/index.php/IDS_OpenIdEngine
Greetings, I'm currently begining to work with neura networks. You could use keras to make layers, torch for process the full dataframe through torch.nn , being these each contig and "follow" each pattern through full fungi Reports. Even if low genome copies are found, you should be able to Isolate the presence of the particular pattern. This would be interesting if you are pursuing an outbreaking finding or hypothesis. Decission forest should be enough to give you a better idea of the composition of your fungi consorcium.
Best regards.
- Badges
- Science method