COMDS manual

MANUAL for COMDS

November 1993

The program implements combined analysis of segregation, linkage and association under an oligogenic model. The program includes a pre-processor list which allows flexible data format, selection and rejection of records, and creation of new variables through transformation.

The COMDS program allows diathesis to be defined for normal individuals, and severity for affected individuals. The program includes probit and logit models. The main reference is Morton et al (1992).

Input data

Data consist of nuclear families where each individual in the family is represented by a record. All of the individuals of a family must be together in the file. A nuclear family may have a father, a mother, children, a pointer to the father, a pointer to the mother, and a pointer to the children. The pointers are optional.

The following are the types of fields in a COMDS data file. Some are always required and others are used only under certain conditions.

Family ID
(required) gives a unique identification to each family in "A" format.

Position
(required) identifies each member of the family.

          __________________________________________
               Value               Position
          __________________________________________
                 1                 father
                 2                 mother
                 3                 child
                 4                 father's pointer
                 5                 mother's pointer
                 6                 children's pointer
          ___________________________________________

Each data set must have either an affection status field and/or a diathesis field. If an AF field is used a LI field must be included.

Affection status (AF)
gives the affection status of an individual as 0 = normal, 1 = affection, 9 = unknown. Blank is read as 9.
Severity (SV)
(optional) specifies an integer j = 1, ......9 corresponding to the jth frequency in the SV control. Unknown severity is indicated by blank or zero and is read as zero. Severity classes must be ordered by increasing severity.
Liability class (LI)
(required) defines situation. Values are i = 1, ....9, corresponding to the ith risk on the LI control.
Diathesis (DS)
(optional) specifies the diathesis for an individual as an integer corresponding to the jth frequency in the DS control or as a quantitative trait. In the latter case if diathesis for an individual is missing, this field should be blank and not zero. Diathesis classes must be ordered by increasing susceptibility.
Pointer relationship (PT)
(required for pointer records and for children's records whose ascertainments are 1A or 1M). Every pointer record must have a pointer relationship field specified in A3 format if a test locus is specified and as A2b format otherwise. More remote relationship may be specified by Abb, where A = 5,6,7,8,9.
Proband (PR)
(required only for incomplete selection). Selection is complete if all children within a family are coded 0 in this field. Selection is incomplete if at least one child within a family is coded > 0, and the ith PI value from the PI control is assigned to this family. Only one value of i is permitted in a given family (children and parents). For families with multiplex ascertainment, 1A or 1M is coded in the pointer relationship field. The program will distinguish between families with and without parent probands in the calculation of the joint likelihood. If PI = 0 for some value of i, only one child may be a proband and the likelihood will be conditional on affection and severity of that individual. In all other cases proband children are assumed to be drawn at random from the severity distribution. Regardless of PI, likelihood is conditional on severity of pointers and affected parents who are probands.
TL
(required only for data with marker locus). This code is a 2- digit integer read as A2 specifying the allele for an individual. The first character is 1-9 and represents the first allele value, and the second character is also 1-9 and represents the second allele. Individuals with missing marker may be coded as 00 or blank.
Job files
Two job files are required, a list job file and a COMDS job file. The list job file governs reading of data, selection and rejection of records, and creation of new variables through transformation. The processed data are written to a file entitled LISTDAT.cp where "cp" is the name of the executable "cp" file prepared to run the program. The COMDS job file allows the user to estimate parameters, to obtain likelihood ratios for two different hypotheses, and to compare likelihoods for different values of theta. If the width of the line exceeds eighty characters, the line is broken, and an asterisk placed at the end of the first half, and the continuation appears on the next line.

Controls for the list job file

These controls are used in the following order: FM, SI, SD, NA, TR, CC.

Table 1   Pointer relations codes for near relatives with marker
          information

Pointer        First and second characters             Code
____________________________________________________________
Sib (of child), selection of multiplex sibships        1A
Sib (of child), selection of multiplex probands        1M
Sib (of parent)                                        1C
Child                                                  1D
Parent                                                 1E
Half-sib                                               2A
Uncle                                                  2B
Nephew                                                 2C
Grandson                                               2D
Grandfather                                            2E
First cousin                                           3A
Granduncle                                             3B
Grandnephew                                            3C
Great grandson                                         3D
Great grandfather                                      3E
Half-nephew                                            3F
Half first-cousin                                      4A
Great granduncle                                       4B
Great grandnephew                                      4C
Great great grandson                                   4D
Great great grandfather                                4E
Child of first cousin                                  4F
______________________________________________________________

FM-control (required)
This control specifies, within parentheses the format by which the raw input data are to be read. There are 5 possible field types: wX to skip w columns; Aw (w = 1 - 6) for alphanumeric specification; Fw.x for numerical variables, where w.x specifies a field of w digits with x decimal places to the right. S and R fields are used in conjunction with the SI/SD control cards. The TR control card considers only the A and F fields as variables. The PT field must be an A3 field, left adjusted. Blanks in F fields are treated as unknown values, not as zeros.
Form (FM)
(format specification) Example: FM (A5, F1.0, 4X, F2.0, F1.0, F1.0)
SI/SD - control
(optional) Selection and deletion. See Morton et al (1983).
NA control
Names of variables (after transformation if necessary) are specified in order by NA. Variables are F and A fields given on the FM control and additional variables created by transformation. Depending on the job, not all variables need be specified. The variables are described in the Input Data section. They are:
```
     ID - family identification
     PO - position within the family
     AF - affection status
     LI - liability (situational) class
     DS - diathesis
     SV - severity
     PT - pointer relationship
     PR - proband
     TL - marker locus

example:  NA1  ID   A5
          NA2  PO   F1.0
```
TR - control
(optional) Transformations. See Morton et al (1983).

COMDS job file

Controls for the COMDS job file: These controls are used in the following order: CB, PI, LI, DS, SV, TL, PA, IT, RA, RP, CC.

CB control
(required) CB is the header to indicate that a COMDS job file is being used.
PI - control
(required only if the PR field is specified). Ascertainment probability for corresponding sampling class.
Form PI (c1, c2, .....cn)
If a child's PR field is coded I, PI(I) will be assigned to that child's family. The maximum number of pi's is 9. The value c = 0 conditions likelihood on severity of the single proband child, affected parents, and pointers. Other conditional likelihoods consider severity of parents and pointers but not proband children.
LI - control
(required only if the LI field is specified). Within a set of parentheses specify the affection risk for each liability class. Form LI(A1, A2, A3 ....An) The maximum number of classes is 9. An individual is assigned to a liability class by his LI field.
TL - control
(optional, required only if there is a TL field). Within a set of parentheses specify the gene frequencies that correspond to coupling frequencies, C1, C2, ... Cn. There may be a maximum of 9 values. Form TL(P1, P2, ....PN)
PA control
(required). The parameter control supplies trial values to the 24 parameters that may be estimated or fixed. The parameters are: V, U, D, T, Q, DM, TM, QM, TH, K, B, BS, S, SM, P, C1, C2, C3, C4, C5, C6, C7, C8, and C9. The default for each parameter is 0, except for V, TM, K, and P whose defaults are 1, and for D, DM and QM whose defaults are .5.
The PA control provides the trial values for the first IT control immediately following it. The trial values for all subsequent IT controls are the final result of the previous IT control. The PA control also provides the parameter values for the first hypothesis in a PA/RA control. Within a set of parentheses specify a series separated by commas of a parameter name followed by an equal sign and a parameter value. Form PA(V=v, U-u, D=d, C1-c1, C2=c2,....C9=c1) PA(V=1, D=.4, TH=.2, Q=.01, C1=.012, C2=.045)
IT control
(optional) An IT control specifies the parameters to be iterated. All other parameters not specified are held fixed. The initial values for a given IT control are the results of the previous control (PA or IT). In the optional first set of parentheses, C = conditional likelihood, J = joint, and M = mating type. If the first set of parentheses is omitted, conditional is the default. In the optional third set of parentheses, the differentiation interval (H) the truncation upper bound (TRUPB), and the tolerance (TOL) may be specified. See Lalouel (1977) the definition of these terms.
```
Form IT(C, J, or M) (P1, P2, P3, ...Pn) (H=h, T=tol)
     IT(J) (V, U, P)
     IT(D, T, Q) (H = .01)
```
The default values for the third set of parentheses are H = .001, and T = .0001. Whenever Q or any of the C's are estimated, the non-iterated C's (not including C = 0 and Ci = 1) are adjusted so that SUM(CiPi) = Q.
RA control
(optional) The RA control card gives likelihood ratios between two hypotheses. The parameter values for the first hypotheses are taken from the last encountered PA or IT control card. The parameter values for the second hypothesis are specified on the RA control . See the PA control for the parameter names and default values.
Form RA(C, J, or M) (V=v, U=u, D=d, etc.) The first set of parentheses is optional. C = conditional likelihood, J = joint, M = mating type. If the first set of parentheses is omitted, conditional is the default.
CC control
(required) This control card terminates the job, and may also be used for comments.

TESTS OF HYPOTHESES

Logistic models for affection status cannot accommodate a continuous latent variable like the polygenic breeding value, and they permit no precise etiological inference if phenotypes of relatives are used as regressors. However, logistic models are readily extended to oligogenic hypotheses with severity and diathesis as ancillary phenotypes.

Segregation analysis

In the absence of a marker locus, segregation analysis attempts to discriminate a major locus from the pseudomultifactorial null hypothesis. The general dihybrid model is saturated and replaces transmission models for tests and subhypotheses. A major locus is supported if the best model is significantly better than the pseudomultifactorial model. This test will fail to detect a major locus with gene frequency near .5. It will tend to assert a major locus (perhaps falsely) if diathesis is quantitative and fails the normal distributional assumption or if for any situational class the risk in first degree relatives is much greater than the square root of the population risk given in the LI control. It is therefore essential that individuals be assigned an appropriate risk.

Linkage analysis

If marker loci are sampled at random, estimates from segregation analysis may be used with coupling frequencies defaulted (ie, assumed equal). Once loose linkage is suspected, segregation parameters should be estimated simultaneously with theta (for detection of linkage) or theta, K for estimation and calculation of sex-factored lods. If tight linkage is suspected on molecular or statistical grounds, coupling frequencies may be estimated simultaneously. On the hypothesis of pleiotropy theta=0 and the coupling frequencies Ci equal 0 or 1 if alleles are homogeneous and possibly monophyletic. Under conditional likelihood the information about coupling frequencies is relatively small, and the best solution with equal coupling frequencies should be sought first. Coupling frequencies should not be estimated if alleles in parents are labelled arbitrarily, as 12x34, and the alleles within a parent are assigned without regard to the major locus so that they are heterogeneous (polyphyletic).

References

Lalouel JM. (1979) Gemini--A computer program for optimization of general nonlinear functions. Technical Report No. 14, Department of Medical Biophysics and Computing, University of Utah.

Lalouel JM, Morton NE. (1981). Complex segregation analysis with pointers. Hum Hered 31:312.

MacLean CJ, Morton NE, Lew R. (1975). Analysis of family resemblance. IV. Operational characteristics of segregation analysis. Am J Hum Genet 27(3):365.

MacLean CJ, Morton NE, Elston RC, Yee S. (1976). Skewness in commingled distributions. Biometrics 32:695.

Morton NE. Segregation and linkage analysis. In Proceedings, 6th International Congress of Human Genetics, Vol. 2. Alan R. Liss Inc., New York. (in press).

Morton NE, Lalouel JM. (1981). Resolution of linkage for irregular phenotype systems. Hum Hered 31:3.

Morton NE, Mi MP. (1968). Multiplex families with two or more probands. Am J Hum Genet 20:361.

Morton NE, MacLean CJ. (1974). Analysis of family resemblance. III. Complex segregation of quantitative traits. Am J Hum Genet 26(4):489.

Rao DC, Morton NE, Lindsten J, Hulten M, Yee S. (1977). A mapping function for man. Hum Hered 27:99.

Williams WR. (1981). Appendix to Thompson MW, Percy ME, Hutton EM. Mutation in the muscular dystrophies. In: Population and Biological Aspects of Human Mutation, (Hook EB, Porter IH, Eds.), pp. 113-116. Academic Press, New York.

Morton, NE, Rao DC, Lalouel J-M. (1983) Methods of Genetic Epidemiology. Karger, Switzerland.

back to linkage software list