menu
Channels
# All channels
view-forward
# General
view-forward
# CLI
view-forward
# Online
view-forward
Team

Difference NCBI Prok - TypeMat

June 1, 2020 at 10:01am

Difference NCBI Prok - TypeMat

June 1, 2020 at 10:01am
Hi, I try to taxonomically annotate an assembles isolate. I ran it as query against Typemat and NCBI Prok online. There is a huge difference regarding the p-values from class level on. I would have expected, that the results are more similar. What would you suggest to use for classification? I would guess if the prediction accuracy is high enough, I can trust that result, but would that mean, I should run my isolate against both or more references and compare all the results? Is there a way to batch download my results in tabular format? Is there a way to parse results from runs against different references? Which result is "most reliable"? From your paper, I couldn't find exactly how you calculate the p-values. Can you shed some more light on this?

June 2, 2020 at 12:14am
Hello The main difference is that TypeMat has genomes from type material, which covers the majority of formally named species, whereas NCBI_Prok only has complete genomes. For classification purposes I would generally recommend using TypeMat.
  • reply
  • like
There is currently no way to batch-download results, but we're working on it. We expect to have this feature available soon.
  • reply
  • like
The estimation of the empirical p-values is described in Rodriguez-R et al 2018, in Supplementary Methods
  • reply
  • like

June 3, 2020 at 7:51am
Hi , thanks for the reply. There is even no way to download the results via curl as far as I can see, right?
  • reply
  • like

June 30, 2020 at 8:29pm
No, this is currently not possible. However, we currently have a new service geared towards high-throughput analyses such as the one you want to use. This way, we can focus on high-quality manual curation in the MiGA Online website. Would you be interested on joining the early-access to this new service?
  • reply
  • like

August 18, 2020 at 11:39am
Hi , yes why not. please contact me via email ulrike.loeber(at)mdc-berlin.de
  • reply
  • like