GRIMS GPID1 Process

From ICISWiki

________________________________________
From: Sackville Hamilton, Ruaraidh (IRRI)
Sent: Wednesday, 2008 February 06 5:48 PM
To: Prantilla, Roniela (IRRI)
Cc: Alcantara, Adelaida (IRRI)
Subject: GPID1 process

Ella,

Is the process below (in blue) the one you were referring to this morning? I think the process is OK as far as it goes. I haven’t checked whether INGER has separated their IRTP GIDs from the root records – that’s an important change to make.

I see 58,112 IRGC accessions have GPID1=0 so we can simply create a new LOCID for them; and 4,097 have a GPID1 that is not used as the GPID1 or GPID2 of any other GID, so we safely use their GPID1. This leaves 55,150 with GPID1 shared by other GIDs.

To choose an appropriate GID to reference the location data:

• For the moment, ignore all accessions where there is an IRTP-IRGC association. We need to complete correcting those associations before we can attach the location data correctly. For all other accessions:
• if a IRGC GID has data on collection location but no GPID1,

• Create a new GID to represent the collection sample
• Set its GLOCN = the LocID of the collection location
• Set its GDATE = the date of collection
• Set the GPID1 of the IRGC GID = the newly created GID

• if a IRGC GID has data on collection location and a GPID1 that is not shared by other GID

• Use that GID to represent the collection sample
• Set its GLOCN = the LocID of the collection location
• Set its GDATE = the date of collection

• if a IRGC GID has data on collection location and a GPID1 that is shared by at least one other GID

• case by case manual inspection of the data to decide whether it is correct to be shared and if so whether it is appropriate to use it to represent the collection sample. If so, use it. If not, ooo-er.

I think we can refine the process a little more by subdividing the last case. As we’ve discussed before, IRRI used to follow the practice of linking landraces sharing the same cv name to a common GPID1 representing the common notional ancestor of that landrace; this is wrong practice (although it is still being followed by some, as I see some new GIDs doing this). We should ignore those GIDs, even when shared by many GIDs. For accessions with a GPID1 pointing to one of these notional ancestors, it is safe to treat them in the same was as GIDs with GPID1=0. The characteristics of these notional ancestors are:

• METHN=31
• GPID1=GPID2=0
• GLOCN=0 or GLOCN=LOCID of a country
• GDATE=0
• All names have NTYPE=6 (in most cases, there is only one name for the GID)
• NLOCN=0 or NLOCN=LOCID of a country
• NDATE=0

About 10,000 accessions would be in this category.

Ruaraidh