BREEDING AND GENETICS - American Society of Animal Science

236 Effect of herd environment level on the genetic
and phenotypic relationship among milk yield, somatic cell
score, and fertility. H. Castillo-Juarez 1 *, P. A. Oltenacu 2 ,R.W.
Blake 2 ,C.E.McCulloch 2 , and E. G. Cienfuegos-Rivas 3 , 1 Universidad
Autonoma Metropolitana, Mexico, 2 Cornell University, 3 Universidad
Autonoma de Tamaulipas, Mexico.
To evaluate genotype by environment interactions of mature equivalent
milk yield (MY), somatic cell score (SCS), and conception rate
at first service (CR), 248,230 first parity DHIA Holstein records from
588 sires in 3,042 herds were used. Herds were classified into low and
high management levels using three criteria. Genetic parameters were
estimated using multiple trait derivative free REML software (MTD-
FREML). Heritabilities and genetic and phenotypic correlations were
consistent regardless of the classification criteria. For low level, heritabilities
for MY, SCS, and CR averaged .232, .101, and .020, while for
high level they averaged .283, .097, and .009. For low level, genetic (and
phenotypic) correlations between MY and SCS, MY and CR, SCS and
CR averaged .234, −.407, and −.228, (−.055, −.174, and −.037), while
for high level they averaged .178, −.304, and −.139, (−.089, −.171, and
−.034). The genetic correlation between low and high management levels
for MY, SCS, and CR averaged .972, .972, and .949. The genetic
correlation between pairs of traits were consistently lower in high than
in low management groups, indicating a genotype by environment interaction.
These changes are all in a positive direction, suggesting that
differences of management between two levels reduces the genetic negative
association between the traits considered.
Key Words: Genetic Parameters, Herd Management, Multiple Trait Linear
Model
237 Acquisition of milk recording data classified by
experts for machine learning. D. Pietersma*, R. Lacroix, and K.
M. Wade, McGill University, Montreal, Canada.
Every month, DHI provides dairy producers with a large number of
test day values related to cows’ milk yield, fat, protein and somatic cell
count. While proper interpretation of these data may produce useful
information to improve management, the process itself is generally time
consuming. A knowledge-based system that could automate parts of this
process and support dairy producers and their advisors would, therefore,
be useful. Since traditional approaches to knowledge-based system development,
based on interviews with domain experts, have also proven to
be very time consuming and costly, alternative, machine learning techniques
such as rule induction and case-based reasoning could be used
to automate the process. This is achieved by learning how to classify
new cases, based on sets of example cases that have been classified by
domain experts. In this project, the use of machine learning to support
the development of a knowledge-based system for the interpretation of
lactation data was investigated. An important phase consisted of collecting
example cases of the interpretation of mature equivalent milk
production and lactation curves by domain experts. In consultation
with domain experts, the different steps in the interpretation process
were determined, such as removal of abnormal cases and classification
of the performance of different groups of cows. In addition, the representations
of example cases were determined for each of these steps.
These include the average performance of a group, an indication of the
variability of the data, reference values and class descriptions such as
high peak or high persistency. A computer program was developed to
show experts example cases and record their classification decisions. By
use of this program, domain experts can classify test day records obtained
from DHI and machine learning techniques can subsequently be
used to transform the knowledge embedded in these example cases in a
format that can later be used in a knowledge-based system.
Key Words: Milk Recording Data, Knowledge-based Systems, Machine
Learning
238 Information technology via the Internet. M. A.
Varner 1 and R. A. Cady 2 , 1 University of Maryland, 2 Washington State
University.
As the Internet has grown in importance and in capabilities as a communication
tool, so have our thoughts on how information technology
might be used in dairy breeding. The two have acted synergistically.
New applications are being developed for one, whenever additional capabilities
are added in the other. The Internet was developed to allow for
remote users to access programs and scientific data on large mainframe
computers. Dairy producers and industry professionals are often in geographically
remote areas due to their career, and they now have much
greater access to dairy breeding information via the Internet. Some examples
include sire summaries that are available at the same time to
everyone on the Internet, as they are to the bull studs. Producers also
have access to sire information from a variety of studs, where individual
bulls can be compared and purchase decisions can be made without
the influence of an on-site sales representative. The technologies have
also allowed for the development of new businesses to provide facilitated
computing environments via the Internet for this kind of cross-stud bull
comparisons. Beyond just the distribution and sales of semen, dairy
breeding is also a process, frequently a collaborative effort among individuals
in various sectors of the dairy industry who are physically remote
from each other. It is often difficult for these individuals to meet in person
for committee or organizational meetings. There are a number of
new information technologies available via the Internet that facilitate
collaborative efforts. Some of those include immediate access to remote
databases, software agents that can carry out assigned tasks, meeting
room or desktop video-conferencing, whiteboard technologies and online
expert systems. Examples of where these technologies might be
useful in existing dairy breeding tasks include committees that need to
meet to evaluate sire summaries, reporting of genetic abnormalities observed,
service to remotely monitor semen supplies, and interpretation
of records. The new technologies also may allow producers to market
their cow information directly to bull studs, who might use it to calculate
value-added information about proofs, bypassing the dairy herd
record processing centers. Trends in information technologies concerning
decision support systems with application to dairy breeding will
be discussed. New technologies often have unanticipated impacts when
used in new domains. Potential area where unanticipated impacts might
develop will also be discussed.
Key Words: Internet, Breeding, Information Technologies
239 A genetic parameter estimate World Wide Web
site. S. Newman 1 , J. McEwan 2 , A. Swan* 3 , L. Brash 3 , and S.
Hermesch 4 , 1 CSIRO Tropical Agriculture, Rockhampton, Australia,
2 AgResearch, Mosgiel New Zealand, 3 CSIRO Animal Production,
Armidale, Australia, 4 AGBU, University of New England, Armidale,
Australia.
The primary consideration in the development of structured breeding
programs is the breeding objective and its predictor, the selection index.
The breeding objective is simply a statement (model) describing the relationship
between various biological traits and income and expense.
Genetic parameters like (co)variances, correlations and heritabilities are
essential to their construction. Other uses of genetic parameters arise as
a function of our ability to model larger and more complex systems and
attempt to integrate information across herds (flocks), breeds, or enterprises.
Examples include multi-breed EBV estimation and modelling
of crossbreeding systems. Given that geneticists do not have access to
the experimental populations they might require to estimate required
genetic parameters (and the costs to maintain them), the ability to access
one central source for these data will be a tremendous utility for
all animal geneticists. A web site has the capability of: Tabulating and
summarising genetic information from major farmed domestic species;
Containing components including breed and heterosis estimates, genetic
(co)variances, heritabilities and correlations; Containing auxiliary information
including age at measurement, fixed effect adjustments, trait
means and method of estimation; Having a nominated person (coordinator)
responsible for each species; allowing electronic submission of parameters;
Providing for older material to be loaded when electronically
submitted by the author or nominee; Allowing access via hierarchical
menus and search functions; Being cost effective.
Key Words: Genetic Parameters, World Wide Web
J. Anim. Sci. Vol. 76, Suppl. 1/J. Dairy Sci. Vol. 81, Suppl. 1/1998 61