Presentation on theme: "Pedigrees wrongly uploaded from Fibos to PMS 2012-08-20 Jesper Nørgaard Welen (SIU) PMS = Pedigree Management System."— Presentation transcript:
Pedigrees wrongly uploaded from Fibos to PMS Jesper Nørgaard Welen (SIU) PMS = Pedigree Management System
Normal backcrosses used world-wide A B A female backross (F) A male backcross (M) B B A/2*B Normal backcrosses recognized all over the world as using the Purdy pedigree representation A*2/B
CIMMYT extra backcrosses (saves ink) A B B male backross (M) A female backcross (F) B A A*2/B CIMMYT-only backcrosses which were decided for the PMS in VAX system around 1990, because they will save ink in printing fieldbooks even if the pedigree string makes confusion possible with the pedigrees of the previous slide A/2*B B//A/B would be normalA/B//A would be normal
Pedigrees wrongly uploaded from Fibos to PMS PCT is the cross type and the values are (F=female M=male S=simple T=Top D=Double) When uploading Fibos to PMS, crosses were given the PCT (cross type) code for backcross or PCT=S for “simple” using as only input the breeders instructions. The upload did not put the relevant selection, only SID=0. In most cases probably the breeder knew if it was a backcross or not, but the PCT was often contrary to the parents given, in other words analyzing the parents would lead to that the PCT field should be different. Breeders or assistants have not detected or complained about this problem (missing SID, wrong cross type). These historic lines where mostly not touched
Interpretation of Fibos pedigrees In the IWIS2 programs were developed and the pmsupdate program will handle the PCT field automatically whenever a cross is updated or created. But if the cross is unchanged, PCT can still remain with a wrong value AB A A*2/B is correct if it is a backcross with SID=0 for both A’s A//A/B is correct if it is not a backcross and SID=x and SID=y are the two selections representing the name A
Pedigrees wrongly uploaded from Fibos to PMS Cross year for these 7,800 F1s are mostly 1989 to 2000, half of them are from the years 1989 and 1990 only Relatively few of these problems are from after IWIS2 was in production from the end of 2000 When a cross is “backcrossed” with two different selections of a cross, it is not a backcross. IWIS2 and IWIS3 respect this. However when the SID is not captured, this means that it will be identified as a backcross Publications often don‘t show the selections, so when implementing in IWIS, the SID=0 is used. A/2*B from a publication will mean that the A and recurrent parent B was most likely both selections (non-segregating material), but we don‘t know them so SID=0 is used
Backrosses represented with the PCT field Suppose we made a rule that a recurrent parent has SID=0 would mean that it is not a backcross. Foreign material however means backcrosses with SID=0 as recurrent parent are frequent, and this rule would change them to not be backcrosses anymore. Therefore this rule can't be implemented in general Making algorithms to produce the same pedigrees in IWIS3 as in IWIS2 has proven quite a challenge. One difficulty is that the data structure for PMS (IWIS2) is fundamentally different to GMS (IWIS3) In IWIS3 there is no backcross field like PCT from IWIS2. Using PCT is a problematic implementation of backcrosses, since the trees must be traversed to figure out if any cross changed backcross status
Updating PCT field is complicated and error-prone update A B A A*2/B is correct C C*2/B is wrong as given by IWIS2 B A A//C/B is correct The dilemma about the PCT field is that it needs to be refreshed for all crosses that are using a given cross for which parents are changed. It is not enough to process the PCT field for the cross that is changed itself (here D), but for all crosses that has D as parent D D
Pedigree curation reveals problem When pedigree curation was done on IWIS3 it was discovered that these discrepancies between IWIS2 and IWIS3 existed. Mostly IWIS3 would show these Fibos F1s as backcrosses, but as simple crosses in IWIS2 Breeders have complained after the change that these should not have been backcrosses and the reason was that the Fibos conversion put parents with SID=0 when this was not accurate, so there is a discrepancy between original breeder‘s knowledge and the way it was implemented
Decision to convert backcrosses This problem I interpreted as being wrong PCT value in IWIS2 for these Fibos F1s and a change of PCT=S simple crosses to backcrosses (F or M) was done on 7,800+ crosses at dates and I took the view that they should have been backcrosses. I consulted with Tpayne & Mgomez & Sgonzales but took this decision to change these from simple crosses to backcrosses knowing it could be reverted. Because this was viewed as correcting “wrong” to “correct”, breeders and assistants were not consulted. The effect in the fieldbooks of these pedigree changes could be seen to be low, but the effect on pedigrees using these 7,800 crosses as parents, was largely unpredictable. If a named cross is used as parent or grandparent, it will only be present in the pedigree if any immediate parents are not named.
Recovering historic hand-written fieldbooks update A B A A*2/B is current A B A The absolute correctness would be obtained by analyzing the old hand-written fieldbooks and recover/recreate which selections would have been involved in these 7,800 crosses. However it is undoubtedly a difficult task and quite likely some selections can't be recovered anymore A//A/B was original before 2011 SID=0 SID=y SID= x A//A/B is restored as before 2011
Alternatives for solving backcross problem If we leave the PCT fields in IWIS2 as they were before 2011, thousands of pedigrees will show differently in IWIS2 and IWIS3, because the flexible pedigrees of IWIS3 are independent of PCT which is not part of IWIS3. The IWIS3 pedigree algorithm deduces correctly what is a backcross and what is not based on the parents alone It is possible to leave status quo as it is because now the pedigrees of IWIS2 and IWIS3 match for these 7,800 crosses (and more that are derived). However that is not the solution I recommend. It might not represent the original breeding the best way The proposed solution is to change the SID=0 to be SID<>0 for the above 7,800 so that the supposed recurrent parent is not the same. SID would represent a dummy selection with no location like CM34-1M-?
Proposed solution update A B A A*2/B is the current pedigree A B A Since the Fibos breeder(?) put PCT field as “simple” that may have been the correct status, except we can't check it because we don‘t have the correct selections anymore. With the above change the pedigree is restored, and consistent between IWIS2 and IWIS3 A//A/B was original before 2011 SID=0 SID= x A//A/B is restored as before 2011 CM3700-F-1M-?
Scripts developed, solution can be reverted The change for these 7,800 crosses to restore the original pedigree can be made with scripts that I have developed and also using the binnacle (change tracking) of IWIS2, and also the pedigree names in IWIS3. It should be automatic and could be finished within a day If there are unforeseen side effects of this change it is always possible to revert it, and even revert the changes made in and The issue here is what is the most correct change to the crosses while keeping consistent pedigrees between IWIS2 and IWIS3