I am Ashok Kumar Sharma pursuing my Ph.D. in metagenomics and systems biology laboratory. I wish to use RAVEN for the draft construction of metabolic network of a genome (* genome is not specific). According to the manual either we can go for blast search using template or can use HMMER.
So I took a genome (contains around 3500 genes), and performed HMMER search using command specified in RAVEN. What I got was around 1000 reactions with respective to ko id hit.
My query is why some of the genes (reactions related to that gene) not present in the draft model and at the same time that genes already classified in the metabolic categories according to the COG classification in ptt file.
Just for an example gi id classified in ptt file like this:
COG0169E shikimate 5-dehydrogenase
The same gi id and reaction is missing in the draft model. And when I searched the gi id in Ko hits given by raven so I found the same reaction at under several Ko files.
Example:
K00014.out:|gi_1| shikimate 5-dehydrog -1442.4 5.3e-25 1
K00015.out:|gi_2| shikimate 5-dehydrog -282.0 4 1
K00020.out:|gi_3| shikimate 5-dehydrog -196.5 4.8 1
and many more hits.
I am not getting, when the reaction is originally present and also the same gene get hits as Ko number, then why is this reaction not included in draft network?
Is there any e-value or score criteria for deciding whether this reaction will be included or not?
This is just an example I have around 500 of these genes for which reactions are missing in the model.
Since, I am using the tool for the first time, I can not go any further without solving it.
Kindly advise me in this regard.
Dear Ashok
First, the KEGG orthology (KO) ids which you aptly mention are just an extension of the identifiers. You may not expect each of them to be mapped. The genes are mapped to the KO terms which are linked successively to the EC#. Please remember that association is transitive. So some of the genes that are associated to the reactions via the EC# only.
Now comes your cut-off, how would you decide whether or not the reaction is optimal to your query. You may have to filter out the KO hits from "conserved" KO assignments above based on probabilistic statistics like e value etc.
Best
Prash
Dear Prashanth,
First of all thank you very much for your kind input.
I agree with your points, Here, Ko extension in the example (I have mentioned above) is the default output of RAVEN. Yes, each ko can not map but here I am considering only those genes which have prespecified metabolic categories in the genome ptt file and for the same genes EC numbers and reactions are also present when I used model seed for draft network reconstruction.
So I was wondering if the gene can be mapped to EC number via model seed and for the same gene in the RAVEN model, EC number and respective reaction is absent whereas gene gets ko hit annotation same as model seed annotation.
Output of model seed is here:
fig|565050.3.peg.3 peg Shikimate 5-dehydrogenase I alpha (EC 1.1.1.25) # AroDI alpha 1700 2558 for rxn01740
I have around 500 genes for which metabolic category specified in ptt file, but all are absent in the model from RAVEN.
How can I include as many as reactions, belonging to those genes, into the model to complete the network?
Thank you.
Pleasure Ashok :)
What you could do is take this to excel worksheet ..data text to columns and filter out those with stringent metabolic categories.
I am not sure of filtering options in Raven though. Probably you could rank them based on enrichment scores and then consider the best hits for a model.
Finally, could the model be perused as a network depends on the threshold you set up. Else use the network as a model again and perhaps you may have to use a training classifier for better prediction.
Best
Prash
- Badges
- Science topic
- Similar topics
- Biological Science
- Molecular Biology
- Biomolecules
- Nucleic Acids
- DNA