Likelihood analysis of segregation distortion
in sperm-typing data for

Grewal RP, Cancel G, Leeflang EP, Duerr A, McPeek MS, Draghinas D, Yao X, Stevanin G, Alnot M-O, Brice A, Arnheim N (1999) "French Machado-Joseph Disease patients do not exhibit gametic segregation distortion: A sperm typing analysis" Human Molecular Genetics 8:1779-1784.

Data


Data are presented from 5 donors who are MJD patients of French descent. For details, see the paper.

For donor 1, we have the following data on 156 single sperm typed at both the MJD1 locus and at the closely-linked marker D14S1050, where A represents the mutant MJD1 allele, a represents the wild-type MJD1 allele, B represents the D14S1050 allele linked to the mutant MJD1 allele, and b represents the D14S1050 allele linked the the wild-type MJD1 allele.


Coamplification data (Donor 1)

----
---b
--a-
--ab
-B--
-B-b
-Ba-
-Bab
A---
A--b
A-a-
A-ab
AB--
AB-b
ABa-
ABab


Total
12
2
4
67
15
0
1
0
1
1
0
0
50
2
1
0


156


For donors 1 through 5, we have the following data on sperm typed at D14S1050, where B represents the allele linked the mutant MJD1 allele while b represents the allele linked to the wild-type MJD1 allele.


Donor 1

----
---b
-B--
-B-b

Total
19
92
95
0

206


Donor 2

----
---b
-B--
-B-b

Total
40
90
91
0

221


Donor 3

----
---b
-B--
-B-b

Total
15
87
91
0

193


Donor 4

----
---b
-B--
-B-b

Total
29
104
98
4

235


Donor 5

----
---b
-B--
-B-b

Total
9
79
83
0

171


Model


We performed a likelihood-based analysis of segregation distortion in the single sperm data using the SPERMSEG program of McPeek (1999) "SPERMSEG: analysis of segregation distortion in sigle-sperm data" American Journal of Human Genetics 65:1195-1197. The specific model fit to these data assumed

  • a common probability (s) of segregation of the mutant MJD1 allele across donors,

  • allele-specific amplification rates that are the same across donors: the amplification rates are denoted by am(A), am(a), am(B), and am(b) for the mutant MJD1 allele, wild-type MDJ1 allele, allele of D14S1050 linked to the mutant MJD1 allele, and allele of D14S1050 linked to the wild-type MJD1 allele, respectively.

  • allele-specific contamination rates that are the same across donors: are c(A), c(a), c(B), and c(b), respectively.

  • To take into account inter-experimental variability, the probability of one sperm deposited is different for each experiment: d1(12), d1(11), d1(2), d1(3), d1(4), and d1(5) for donor 1's 2-marker data set, donor 1's 1-marker data set, donor 2, donor 3, donor 4, and donor 5.

  • The probability of two sperm deposited is the same across donors (d2).

  • The recombination fraction between MJD1 and D14S1050 was assumed to be 1%.


    Parameter Estimates (with 95% confidence intervals)


    Parameter

    s

    d1(12)
    d1(11)
    d1(2)
    d1(3)
    d1(4)
    d1(5)

    d2

    am(A)
    am(B)
    am(a)
    am(b)

    c(A)
    c(B)
    c(a)
    c(b)
    Estimate

    0.496009

    0.923240
    0.940795
    0.847629
    0.956031
    0.908047
    0.982358

    0.000000

    0.777733
    0.981824
    0.974617
    0.946011

    0.000000
    0.000000
    0.018374
    0.011326
    Lower Limit

    .461
    Upper Limit

    .531


    Details of the Analysis


    A Monte Carlo goodness-of-fit test using SPERMSEG did not indicate misfit of this model to the data (p-value .3). A likelihood-ratio chi-square test comparing the above model to the model in which the segregation parameter is set equal to 1/2 does not indicate significant segregation distortion (p-value .82). Significant inter-experimental variability showed up in the parameter representing probability of 1 sperm deposited (d1). A Monte Carlo goodness-of-fit test of the model in which all d1 parameters were equal showed significant misfit (simulated p-value < .05). The mutant MJD1 allele showed reduced amplification, as is sometimes the case with large repeats, which could potentially result in a biased estimate of segregation probability and hide any segregation distortion in favor of that allele. However, the amplification rates of the two alleles of the linked marker were very close, with the amplification rate for the allele linked to the mutant MJD1 allele actually being slightly higher (not significantly different p-value .29). Thus, using the linked marker, we do not expect the estimate of segregation distortion to be biased, particularly not in the direction of hiding segregation distortion in favor the the mutant MJD1. We do not detect such segregation distortion in these data.