This question was also asked on Biostars 3 months ago, i.e. a long time - so its perfectly reasonable for the question to be asked here

I started studying bioinformatics and I enjoy it a lot (I'm a software developer with around 15 years of experience but mostly in backend, scripting, architecture, and security) and I want to do, as a project, a database regarding natural compounds. When I saw that on NCBI there are a lot of plant genomes I thought, that, with software, somehow we can "guess" (probabilities, not certainties) what a plant can contain or not.

I'm wondering, if, by bioinformatic means only, it is possible to "extract" genes and, after that proteins and natural compounds from plant genome data that is available on NCBI (please see https://www.ncbi.nlm.nih.gov/datasets/genome/GCA_029618835.1/ as an example)? Example: just by processing the data from the link above we can get to Flavoxanthin or any other substance.

I was wondering if a kind of "de-novo" annotation can be made (if that's the correct term).

What other useful information from a chemical compound perspective can be found from digging on the genome itself?

Thank you!

  • $\begingroup$ Please edit the question to limit it to a specific problem with enough detail to identify an adequate answer. $\endgroup$
    – Community Bot
    Commented Jun 13, 2023 at 12:37
  • 1
    $\begingroup$ Why do you want to do this? There are lots of different tools; it would help to be more specific in order to make the answers actually useful. $\endgroup$
    – gringer
    Commented Jun 14, 2023 at 21:08
  • $\begingroup$ @gringer .. you are totally right, I put full details on another platform when I saw that the interest here was almost none: biostars.org/p/9566425 $\endgroup$ Commented Oct 8, 2023 at 13:22
  • $\begingroup$ Thanks for the additional information, but your additional details still describe a generic problem, and are therefore unlikely to get good answers. What is the specific problem that you are trying to solve? $\endgroup$
    – gringer
    Commented Oct 8, 2023 at 20:20

2 Answers 2


This sounds like you are wanting to perform a machine learning calculation. However, for this question leveraging KEGG, Kyoto Encyclopaedia of Genome and Gene and building an SQL database around it is much more effective. It represents around 60% of all functional genes. KEGG is the best gene annotation out there IMO, particularly for flavonoids etc... Fantastic and all their linking pathways, everything links to everything, just amazing. KEGG represents an enormous volume of work, much of it being manually crunching publications for each gene.

The entire database of KEGG keys is downloadable. Connecting a gene to a class of function is easy via KEGG the relationships are defined its just SQL-ing the whole thing together (querying all the tables). Again, it's totally and completely brilliant. Sure they'll be stuff out there has SQL-ed KEGG key tables, but if your a developer its better to build it independently - this way you're certain whats in there and what its doing.

However, linking raw sequence data to a KEGG gene code is more challenging. lastKOALA and GhostKOALA is ... yeah well an approach no more. There's a main stream approach but with current limitations in implementation. KEGG however isn't limiting, it's linking a gene to a KEGG id is the challenge. Getting hold of the raw sequence for KEGG would be great, fantastic in fact, but I'm yet to find it.


I've used Augustus stand alone (https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1160219/) to find genes in a given eukaryotic genome. I agree you can then use things like KofamSCAN to categorize the genes against the KEGG database. If you want to compare against the NCBI database, diamond blast (https://github.com/bbuchfink/diamond) is probably the quickest way to go :)


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.