University of Michigan Center for Statistical 


LAMP -- Program Options

Input Files and Basic Parameters

-d datafile
Selects input data file, in MERLIN format.
-p pedfile
Selects pedigree file, with genotype, phenotype and family structure information.
-f framework map file
Selects map file for flanking markers. See section on Input Format for further details. This file is optional.
-c candidate map file
Selects map file for candidate SNPs and locations. See section on Input Format for further details.

Additional options

Genetic Model Constraints

LAMP can optimize an unconstrained genetic model or can be direct to use an additive, dominant, recessive or multiplicative model. When accurately specified, constrained models reduce the number of parameters to be estimated and can increase power. In addition, you should always set the prevalence to a value appropriate for your disease.
Assumes that the heterozygote genotype has an intermediate penetrance in relation to the two homozygotes.
Assumes that an effect on penetrance is only seen for homozygous genotypes.
Assumes that an identical effect on penetrance is seen for individuals who carry one or two copies of the risk allele.
Assumes that the penetrance is multiplied by a constant factor with each additional copy of the risk allele.
This gives an unconstrained genetic model.
This option fixes the prevalence at the user-specified value. The default prevalence is set at 0.05.
Note that the dominant and recessive models lead to identical likelihoods since, for a two-allele disease model, finding one dominant allele implies that a recessive allele is also present!

Parameter Constraints

Depending on your sampling strategy, not all parameters may be identifiable. LAMP uses some simple rules to decide what parameters can be estimated in each sample, but it is sometimes wrong. You can override LAMP's automatic choice with one of the options.
This option specifies that all linkage information should be ignored. When the option is selected, LAMP will only fit models that imply direct association between each evaluated marker and the trait. The option is always appropriate for samples of unrelated individuals and for trio data, but can also be useful in other samples that include very little information about genetic linkage. We commonly use it for the analysis of genome-wide association scans. The option was formerly known as --unrelateds-and-trios.
Ascertained sibling pair samples that include only affected sibpairs, only unaffected sibpairs, or only discordant sibpairs contain limited information on genetic linkage. When dealing with these samples and no genetic model is specified, LAMP fixes the allele frequency for linkage models to be the square root of the penetrance. This results in no loss of power, but makes the model identifiable.
Do not place any constraints on the allele frequencies and penetrance parameters
Apply some simple rules to try and automatically classify the available sample

Optimization Strategy

When fitting a genetic model, LAMP maximizes a likelihood numerically. There are a number of alternative optimization strategies to choose from and it is also possible to restart each of the strategies from different randomly selected parameter settings.

This is a quasi-Newton method that uses partial derivatives calculated numerically.
This is the default optimizer based on the simplex method. It is quite robust and seems to perform well in most settings.
This optimizer implements a stochastic search procedure, as described by Storn and Price (1996). It appears to perform well, but is quite slow.
--rounds n
At each stage, select n different starting points and repeat the optimization procedure. This is a good final check for the most exciting findings.
--precision x
Numeric optimization will stop when the relative improvement in the log-likelihood is smaller than x. By default, x is set to 10-8.


These options help tune the performance of LAMP and can be used to tweak memory and computing time requirements.

--maxbits n
The maximum complexity of pedigrees that LAMP will try to analyze. Lowering this value will result in discarding of large pedigrees, while increasing it will result in attempts to analyze more complex pedigrees and longer run-times. Bit complexity refers to the depth of the gene-flow trees used to represent the pedigree internally.
By default, LAMP buffers intermediate results to try and speed up calculations. This increases memory use and may not be very helpful in datasets that include only a few large pedigrees. To disable buffering, add the --buffers:OFF option to the command line.
The causality test is most useful after identifying one more associated SNPs. This option can be used to disable the test in first pass analyses, therefore saving time.

Output Files

There are two options to control generation of output files:

This option can be used to suppress generation of detailed output files and therefore preserve diskspace. When this option is not used, LAMP generates a series of files to summarize estimated model parameters. (By default, these are called lamp-base.out, lamp-linkage.out, lamp-direct-association.out and lamp-indirect-association.out.)

--prefix string
This option can be used to control the name of detailed output files. It is convenient when LAMP is run multiple times in the same directory.


University of Michigan | School of Public Health | Abecasis Lab