Table III
Strain:
|
|
|
Curly
|
|
|
tra2ts2
|
|
|
|
|
|
|
|
|
|
|
Prediction:
|
|
f
|
m
|
|
f
|
m
|
Truth:
|
percentages:
|
|
|
|
|
|
|
|
|
|
Curly (18°C):
|
female
|
|
66.67
|
33.33
|
|
0.00
|
0.00
|
|
male
|
|
17.78
|
82.22
|
|
0.00
|
0.00
|
|
|
|
|
|
|
|
|
|
|
|
tra2ts2 (18°C)
|
female
|
|
0.00
|
0.00
|
|
100
|
0.00
|
*
|
|
male
|
|
0.00
|
3.45
|
|
10.34
|
86.21
|
|
|
|
|
|
|
|
|
|
|
|
Table III. Truth contingency table for 18°C controls of
tra2ts2 temperature-shift sex discrimination experiment on
Drosophila melanogaster. This table documents a typical control group
result of this f2-hybrid cross in which the objective is to sort out the
tra2 genetic females (*) in order to
assess the sexual phenotype that they are expressing. We demonstrate the
effectiveness of using wing landmark discriminant functions in helping
accomplish the goal. The truth in
this table is represented by the 'apparent' Curly phenotype
(tra2 segregants are non-Curly) and Barr-eye-of stone tag on
males determines sex of the individuals.
Independently, a prediction of sex
and Curly phenotype was made using four discriminant functions based on
the analysis of the wings' landmark coordinates. The most important result is
that when a truth assignment to Curly is made it is invariably correct
(i.e. the probability of a false positive is vanishingly small). However
sometimes the Curly phenotype is leaky and appears close to the
non-Curly wild type (i.e. the probability of a false negative is not
zero). Luckily, the Curly phenotype is also very distinctive as
recognized by its discriminant function. Occasionally a presumed tra
(i.e. by its apparent non-Curly appearance by visual inspection)
is predicted to be Curly by the discriminant function. These wings
(the 3.45% of tra males in row four of the table) are removed from
the analysis and assumed to have been Curly visual non-penetrants.
This two-level sorting of the experimental strain is relied upon during
temperature shift experiments to sort out potential false positive tra-2
segregants. Using this double sorting it is seen that the control group
of interest in the above table,
tra2ts2 (18°C)
females,
behaved perfectly with 100% of the row's individuals being predicted to be
tra2 females.
Last update: 97/01/01
J. G. Kunkel