Hello,
Please could "indirect" and "unknown" be added
as "interaction type"s in the controlled_vocab.? (I
did email this around a couple of months ago, so
apologies if this has already been addressed and I
have missed an email).
http://psidev.sourceforge.net/mi/rel25/data/psi-
mi25.obo
From literature or data vendors one can not always
make a "confident" call on whether an interaction is
is direct or indirect, so that is why unknown is
useful. Having some general terms like direct,
indirect and unknown helps people, without having to
create yet another attribute. "direct interaction"
(MI:0407)
We have been using them but just used MI:0190
(interaction type) with our additional terms. Even if
it will only be in a future controlled_vocab. listing
that is fine, but it would be useful to know what the
MI:numbers will be.
Alternatively, could there be a field for
directionality of interactions? so that it is quick
to see whether the evidence points to an interaction
being direct, indirect or unknown(unsubstantiated).
Thanks a lot,
Peter
Bioformatics Science and Technology
GSK
Logged In: YES
user_id=653048
Although we perfectly understand the need of directionality
distinction we do not see use case for the term 'indirect
interaction'.
When a interaction is analyzed if there is not enough
evidence to state that the interaction is direct we use the
interaction type 'physical interaction' MI:0218. This term
covers your 'unknown' status and is defined as "Interaction
among molecules that can be direct or indirect". When
proteins are purified or the overall experimental setting is
appropriate the interactions are promoted to 'direct'. I do
not see cases when you can assume an interaction is
'indirect'. In my opinion an interaction between protein A
and protein B is prove to be indirect only when a 'bridging'
protein C is observed/identified (in this case the
interactions A-C and B-C would be reported as 'direct'
interactions) otherwise it remains an interaction with
'unknown directionality' thus a physical interaction in
PSI-MI terms. A classical example could be the TAP complexes
that certainly contain a mixture of direct and indirect
interactions but are annotated as 'physical interaction' as
the data do not provide such detailed information.
If you do not agree, please send me examples of assays that
provide interaction you can assume to be 'indirect' or
papers that contains interactions you would annotate as
'indirect'.
For now consider the following mapping for interaction data
from public databases:
-direct : 'direct interaction' MI:0407 and all direct
interaction children terms
-unknown : 'physical interaction' MI:0218, 'colocalization'
MI:0403, 'genetic interaction' MI:0208 and all genetic
interaction children terms
-indirect interaction : Non Available
Luisa