The University of Arizona

FPC Simulation with error  

Home | Search | FPC | Contact Us

This simulation compares different coverages and error. It also compares HICF and agarose. The results show that HICF is better than agarose even with error. The results also show the benefit of reducing your error.

Type   Set   Cutoff
Coverage
Error
#Ctgs
Score
#Qs (#ctgs w/Qs)
HICF     1   1e-40






10x
6%
24
0.892
0(0)

10x
12.5%
46
0.802
39(11)

20x
6%
2
0.874
0(0)

20x
12.5%
8
0.710
328(7)
HICF     2    1e-50






10x
6%
43
0.886
0

10x
12.5%
79
0.797
13(9)

20x
6%
9
0.874
1(1)

20x
12.5%
17
0.724
106(10)
Agarose  2  1e-12






10x
6%
163
0.959
0(0)

10x
12.5%
211
0.921
8(8)

20x
6%
67
0.928
0(0)

20x
12.5%
147
0.872
55(33)
The same set of clones were used for a given set, except the 20x has an additional 10x clones. So the same clones are used for the HICF Set2 and Agarose Set2.
Though the number of contigs for the 20x/12.5% error is less than for the 10x/6% error, the number of Qs is much higher, implying an increase in the problem within the map. These problems will lead to more time for manually editing the map and increased difficulty in selecting a MTP (see MTP simulation). The take home message from this simulation is that if you reduce your error, you can reduce your coverage. Moreover, the amount of overlap between clones (i.e. the endpoint coordiantes) will be closer to correct. Note, they can never be exact since there in not enough information in the bands, but the less error in the fingerprints, the less error in the overlap. Reduce your error!!

For HICF, we typically get 12.5% error per clone. For Agarose, it really depends on how good the band-calling is. We had exceptionally good band-callers for the maize agarose fingerprints and had an average <6% error per clone. To test your error, do the following: (1) refingerprint a plate of clones, (2) for the second plate, give the clones the same name but a different gel name, and add them to your fpc database (Update .cor)  (3) On the Project Window, select Search, then select "Find clones with mult gels > cutoff". Iit will list all the clones where the two gels do not match below the given cutoff. At the end of the output, the final statistics will be displayed as the following example:

  0  51689 17.839
 1  79004 27.265
 2  60871 21.007
 3  40494 13.975
 4  25598 8.834
 5  15881 5.481
 6   9895 3.415
 7   6328 2.184
IBM Bad 2354 Good 12394 Fp 10.49 (7,1e-10, min 0, max 32767)
This says that 17.8% of the bands for this dataset are exactly the same, 27% have a difference of 1, etc. The list goes up to 7 since that is the tolerance. All bands that do not match with another at the tolerance are considered a Fp (false positive). There are 10.49 Fp in this set, which is 5.25% error per clone. Out of all the clones with multiple gels, 2354 were below the cutoff and the rest were good matches.

The above simulations were performed with one Build. On a real dataset, we then perform automatic end joining, which further reduces the number of contigs without manual merging. See the Automerge demo.

Email Comments fpc@agcol.arizona.edu

 

 

 

Last Modified Thursday February 14, 2008 10:45 AM and 01 seconds