Nemaline myopathy

Abstract We describe a large genealogy data base, which can be searched by computer, of 295,095 Amish and Mennonite individuals. The data base was constructed by merging our existing Anabaptist Genealogy Database 2.0 containing approximately 85,000 individuals with a genealogy file containing approximately 242,000 individuals, kindly provided by Mr. James Hostetler. The merging process corrected thousands of inconsistencies and eliminated hundreds of duplicate individuals. Geneticists have long been interested in Anabaptist populations because they are closed and have detailed written genealogies. The creation of an enlarged and unified data base affords the opportunity to examine inbreeding trends and correlates in these populations. We show the following results. The frequency of consanguineous marriages shows steady increase over time and reached approximately 85% for individuals born in 1940-1959. Among consanguineous marriages, the median kinship coefficient stayed stable in the 19th century, but rose from 0.0115 to 0.0151 in the 20th century. There are statistically significant associations (p

KEY WORDS: AMISH, MENNONITE, GENEALOGY, INBREEDING COEFFICIENT, KINSHIP COEFFICIENT, ISOLATED POPULATIONS

In the 18th and 19th centuries several Anabaptist populations fled religious persecution in central Europe. Many of these groups, including Amish, Mennonites, and Hutterites, settled in North America benefiting from more religious freedom and more available farmland. These Anabaptist groups have long been of interest to geneticists because they have remained mostly closed and inbred, and because they maintain detailed genealogies (McKusick 1978). For example, the Lancaster Mennonite Historical Society in Eastern Pennsylvania has a library housing thousands of genealogy books. Other Anabaptist genealogy resource centers exist in Ohio and Indiana. It seems that much more human effort has gone into creating new genealogy books with a special focus (e.g., one surname, one county, descendants of one immigrant) rather than into merging and reconciling existing sources.

The term "Anabaptist" literally means "rebaptizer" and is used to refer to a Christian movement that arose in central Europe in the first half of the 16th century. Adherents support adult baptism, pacifism, and separation of church and state. Among the large Anabaptist groups existing today are Mennonites (who were originally followers of Menno Simons), Amish (originally followers of Jakob Ammann who split away from the Mennonites at the end of the 17th century), and Hutterites (originally followers of Jakob Hutter). There have been further subdivisions within these Anabaptist communities and others. Amish and Mennonites emigrated to North America in multiple waves in the 18th and 19th centuries (Gingerich and Kreider 1986). The Hutterites migrated through Europe between the 1500s and late 1700s, when they were granted religious freedom in Russia; when that freedom was taken away in the late 1800s they began emigrating to the northern and western parts of North America (Ober et al. 1999).

In 1996, we set out to make a computerized data base of a single Lancaster County (Pennsylvania) Amish genealogy entitled "Fisher Family History" (Beiler 1988), which is focused on the descendants of Amish pioneer Christian Fisher (1757-1838). We will abbreviate this genealogy source as FFH. We selected this book because it was by far the most widely used for studying the living Eastern Pennsylvania Amish. Our initial data base was called the Amish Genealogy Data base (AGDB 1.0) and was described along with the query software PedHunter in Agarwala et al. (1998). The name of the data base was changed to Anabaptist Genealogy Database 2.0, but retained the acronym AGDB, when we merged in a second genealogy source entitled "Amish and Amish Mennonite Genealogies" (Gingerich and Kreider 1986), as described in Agarwala et al. (1999).

The principal goal of constructing AGDB has been to assist medical geneticists in automatically and systematically constructing pedigrees for disease gene hunting. Pedigrees constructed using AGDB version 1.0 or 2.0 have been used by us to hunt genes for McKusick-Kaufman syndrome (Stone et al. 1998; Stone et al. 2000) and Amish nemaline myopathy (Johnston et al. 2000) and by others to hunt genes for hypertension (Hsueh et al. 2000a) and diabetes (Hsueh et al. 2000b). Several other disease studies using pedigrees from AGDB are under way. As described in Agarwala et al. (1998), access to AGDB is provided to other investigators who are studying the genetics of North American Anabaptist populations. Because pedigree construction is one of the earliest steps in disease gene hunting, several years may pass between the time an AGDB query is done and the results are published.

While our initial usages of AGDB 1.0 were successful in assisting disease gene hunts, it soon became apparent that merging in more sources would greatly increase the coverage of AGDB. Merging genealogy resources can create new ancestor-descendant paths not present in either source alone. We quantified the positive effects of merged paths in Agarwala et al. (1999) based on increases in apparent inbreeding coefficients and elimination of apparent founders. The source "Amish and Amish Mennonite Genealogies" (henceforth abbreviated as AAMG) was chosen second because it is comprehensive in its coverage of the Amish immigrants and their descendants through the mid-19th century. Therefore, it allows us to trace back pedigrees to dozens of founders other than Christian Fisher.

In this paper we announce the merger of a third and even larger genealogy source into AGDB to create version 3.0. AGDB 3.0 contains 295,095 individuals, making it roughly 3.5 times the size of AGDB 2.0. The new genealogy source is a computer file maintained and kindly provided by Mr. James Hostetler. The new source contains approximately 242,000 individuals and is focused on Amish and Mennonite populations in the Midwestern United States. Because the geographic emphasis is different from our previous sources, there are tens of thousands of individuals in AGDB 2.0 who are not in the Hostetler file. We have already used the increased coverage in AGDB 3.0 to answer some pending queries from users of AGDB, which could not be answered using earlier versions. The Hostetler data includes births through 1999, and we also incorporated an update to FFH, from Miss Katie Beiler, that includes births through 1999. However, the number of births decreases steadily from 1995 through 1999, indicating that a lot of recent births are missing. It is usually the case that within a nuclear family, all live births and even some stillbirths are recorded, but the compilers of the different sources make rather different choices for which nuclear families to record.

Mr. Hostetler is an avid genealogist who collects Anabaptist genealogy data from over 150 sources, including genealogy books, church directories, and personal correspondence. As can be seen from Figure 1 and our description of the merging process, his data include substantial amounts of information from books such as AAMG and FFH, but are still missing much of the material from those sources. Mr. Hostetler's project is active and ongoing, and the version we received from him reflects the data accumulated through 1999 and in the first weeks of 2000.

The principal effort to create AGDB 3.0 was in identifying which individuals in the new Hostetler source match individuals in the FFH or AAMG. In this process thousands of inconsistencies among the three original sources were repaired and hundreds of duplicate entries were removed. This process involved systematic computerized listing of candidates combined with human checking. Since it is very common among Amish and Mennonites for several individuals to have the same name, it is difficult to identify duplicates within a single source.

The creation of a large genealogy data base such as AGDB 3.0 affords the opportunity to examine long-term trends in inbreeding and its correlates. In 1987, Khoury et al. (1987a, b, c, d) published a large-scale epidemiological study of the Lancaster Amish based on a 1972 version of FFH (Egeland 1972). In the present paper we reexamine some of the conclusions of that study based on a data base that is more than six times as large. We also investigate possible associations between inbreeding and family size and birth intervals raised in an intriguing paper of Ober et al. (1999) analyzing a Hutterite population. In that paper the authors suggested that higher levels of inbreeding may be associated with reduced family size and longer interbirth intervals, contradicting several earlier studies.

We also investigate whether there is an association between mother's inbreeding and giving birth to twins. There have been dozens of studies (e.g., Parisi et al. 1983; Philippe 1985; Meulemans et al. 1996) suggesting a genetic component to twinning. Recently, Al-Hendy et al. (2000) reported a pedigree in which there is an association between twinning and homozygous mutations of the follicle-stimulating-hormone receptor, suggesting that there might be some recessively acting genes associated with twinning.

The extension of AGDB to include the Hostetler data and to more than triple its size should provide substantial added benefit to geneticists hunting disease genes in Anabaptist populations. The enlargement will be especially helpful for studying the populations in the Midwestern United States because this is the focus of Hostetler, and because AGDB 2.0 had very few living individuals west of eastern Pennsylvania.

Materials and Methods

Genealogy Parsing and Merging. The Hostetler data was provided to us as an ASCII text file in GEDCOM format, which is commonly used by genealogists. Because of the standard format, parsing the file into tables for the relational data base was relatively straightforward as compared to FFH and AAMG (see Agarwala et al. [1998] and Agarwala et al. [1999] for details of parsing FFH and AAMG, respectively). Merging of two genealogy sources requires matching individuals present in both sources. When an individual in one data base is matched with more than one individual in the other data base, we usually detect a duplicate in the latter data base. Duplicates usually have one copy listed only as a child and the other copy listed only as a parent. Therefore, coalescing duplicates can extend paths of descent, which in turn could result in matching more individuals present in both data bases. Therefore, merging two data bases requires several iterations of matching individuals between data bases and deleting duplicates within a data base, until no new matches or duplicates can be found. We merged the Hostetler data with AGDB 2.0 in 16 iterations with the vast majority of matches between the two sources coming in the first 3 iterations. In the first iteration, we considered individuals with known birth and death dates and matched ones that had same birth and death dates; cases where multiple distinct individuals had the same birth and death dates were checked by eye. In the second iteration, we looked at the notes kept about the source of information for the entry in Hostetler data and considered individuals whose information came from AAMG. Hundreds of problematic cases were checked by eye. In the third iteration, we considered individuals with known birth date and unknown death date or known death date and unknown birth date. We matched individuals that had the same name and for whom the known date agreed. Iterations 4-16 considered nuclear families containing individuals matched so far and considered extending the matches to children and parents not yet matched. All the candidate matches in iterations 4-16 were double-checked by eye. Each iteration included coalescing duplicates and correcting inconsistencies.

Measures of Inbreeding. Two commonly used measures of inbreeding or consanguinity are the inbreeding coefficient and the kinship coefficient. The kinship coefficient of a couple (P, Q) is the prior probability (based on pedigree structure alone) that, for an arbitrary stretch of autosomal DNA, a randomly chosen allele of P and a randomly chosen allele of Q are inherited "identical by descent" from the same common ancestor. The inbreeding coefficient of an individual R is the same as the kinship coefficient of the parents of R. See Weir (1996, 204-208) for more information. We used PedHunter (Agarwala et al. 1998) to compute the kinship coefficient of the spouses in each marriage and to compute inbreeding coefficients for the other analyses.

Consanguineous Marriages. It is useful to track consanguineous marriages (i.e., those with kinship coefficient > 0) over time to assess whether the risk from recessively acting disease susceptibility genes may be increasing. Closed populations may respond to the disease risk from consanguinity by discouraging marriages between close relatives. Therefore, the fact that a population is closed does not imply that all measures of inbreeding necessarily increase over time. We identified all marriages that produced at least one child. We restricted attention to these marriages because the vast majority of marriages with apparently no descendants represent branches not included in the genealogy. Moreover, it is only in the families with children where the effects of inbreeding get passed to subsequent generations. We summarized the marriages by 20-year cohorts of birth year of the mother. We used 20-year cohorts for consistency with Ober et al. (1999). We preferred cohorts by birth year instead of marriage year, because in the sources the marriage year is missing far more often than the birth year.

Family Size. To evaluate a possible association between family size and inbreeding we counted the number of children in each family and calculated the mother's inbreeding coefficient. We subdivided families according to zero inbreeding vs. nonzero inbreeding (to test for association at the low end) and according to bottom 75% and top 25%, exactly as done in Ober et al. (1999). We tested for association with the mother's inbreeding coefficient, so as to directly compare with the results of Ober et al. (1999); we could have instead tested for association with the parents' kinship coefficient or the father's inbreeding coefficient. Our most recent cohort ends with mothers born in 1959, since they are unlikely to bear more children. For the period 1990-1999, 3.58% children born are such that the difference between the mother's birth year and the child's birth year is more than 40. Each pair of groups, within each cohort, was compared by a t-test and by a Wilcoxon rank sum test. These and all other statistical tests were carried out in S-Plus (Becker et al. 1988).

Birth Intervals. To evaluate a possible association between birth intervals and inbreeding we counted the number of months between first and second births in each family, where this information could be discerned. By using only one birth interval in each family, we avoid a complication that Ober et al. (1999) had in using multiple birth intervals in the same family. We again subdivided families according to zero inbreeding vs. nonzero inbreeding and according to bottom 75% and top 25%, exactly as done in Ober et al. (1999). Each pair of groups, within each cohort, was compared by a t-test and by a Wilcoxon rank sum test.

Apparent birth intervals can be affected by poor recording of stillbirths or of infants who died shortly after birth. So far as we can tell, none of the data sources has a bias against reporting infants who died shortly after birth. Some stillbirths are recorded, but they seem to be underrepresented.

Early Death. We identified all individuals who died before their first or 20th birthdays. We used a case-control design to match the individuals who died early ("cases") to individuals who did not ("controls"). Each case of early death was matched to five randomly chosen controls with the same birth year. Sampling of controls was done without replacement. To keep the design simple, we compared how many cases and controls had inbreeding coefficient 0 and > 0 using a chisquared test in each birth cohort.

Twinning. We identified 2703 multiple births out of 271,437 births for a 1% twinning rate, which is very consistent with the rates of twinning collected by Bortolus et al. (1999). There were 28 cases of triplets, and no cases of quadruplets. We assessed a possible association between twinning and mother's inbreeding using a case-control design. "Cases" are mothers of twins such that the mother's birth year is known and is in the interval 1800-1959. For each case we attempted to find a matching "control" with the same birth year and the same number of birth events. A birth of twins or triplets counts as a single birth event, so that a "case" mother with four nontwin children and one pair of twins would get matched to a "control" mother with five nontwin children. When there were multiple possible controls, we selected one at random without replacement. We matched 2365 mothers of twins, born between 1800 and 1959, to controls. Using 20-year cohorts we compared the inbreeding coefficients of cases to controls by a Wilcoxon rank sum test.

Results

Genealogy Summary. The total number of (apparently distinct) individuals in AGDB 3.0 is 295,095. Figure 1 shows a Venn diagram of how many individuals are in each combination of three sources. The newly added Hostetler data contains 32,102 (= 15,424 + 15,569 + 1109) individuals already in the union of FFH and AAMG. This is a substantial overlap, but less than 40% of the total number of individuals in those two data sources. Therefore, we would expect the union of the three data sources to be substantially more useful than any two together.

The incomplete overlap seen in Figure 1 was expected because of the different foci of the three sources: FFH focuses on Lancaster County; AAMG focuses on early settlers; Hostetler focuses on the Midwest (non-Lancaster) US population. The numbers shown for FFH and AAMG in Figure 1 differ from those reported for AGDB version 2.0 (Agarwala et al. 1999) for two reasons. First, in late 1999 we received and incorporated an update to FFH containing several hundred births. Second, the process of merging the Hostetler file identified 79 duplicates in FFH and 31 duplicates in AAMG with one individual duplicated in both FFH and AAMG. We also identified 524 duplicates in the Hostetler file, and these are not included in the counts in Figure 1. Many of the duplicates are simply two entries with the same name in one source that cannot be convincingly identified as duplicates without external evidence, such as another source. The ability to systematically identify duplicates and correct other such entry errors supports the utility of processing the raw data automatically with computer software.

Consanguineous Marriages. The kinship coefficients of marriages over time are summarized in Tables 1 and 2. There is a steady increase in the fraction of consanguineous marriages from 139/360 (39%) in birth cohort 1800-1819 to 9163/11279 (81 %) in birth cohort 1940-1959. The same trend was observed in Khoury et al. (1987a), but the authors of that study underestimated the fraction of consanguineous marriages in the 19th century. Table 2 shows a perceptible steady increase in the amount of inbreeding for couples married in the 20th century. For example, the median kinship coefficient increases from 0.0115 for mothers in the 1880-1899 birth cohort to 0.0151 for mothers in the 1940-1959 birth cohort. This increase is evident only for marriages occurring after 1950 in the data in Khoury et al. (1987b, Table 1). In our study, trends dependent on calculation of inbreeding become evident in earlier birth cohorts, and we get a higher estimate for the fraction of consanguineous marriages in the 19th century due to presence of the Amish immigrants and their children recorded in AAMG. The time lag may also reflect basic differences in the Lancaster and Midwestern populations.

Family Size. Table 3 shows a very strong positive association between inbreeding and family size, especially in the 20th century. These data are consistent with many earlier studies on closed populations (see, for example, Schull and Neel [1972] and references therein), but contradicts the data on Hutterites in Ober et al. (1999).

Birth Intervals. Table 4 shows the mean birth interval between first and second child depending on the mother's inbreeding. There is a very strong association between shorter birth intervals and higher inbreeding, especially in the 20th century. This finding is consistent with the data showing higher family sizes in closed populations, but contradicts the data in Ober et al. (1999), where the authors suggest that higher inbreeding is associated with longer time to pregnancy.

Early Death. The relationship between early death and inbreeding is summarized in Tables 5 and 6. For birth cohorts through 1919, there appears to be no relationship, presumably due to other common, nongenetic causes of death. For the birth cohorts 1920-1939 and 1940-1959, there is a statistically significant (positive) association between inbreeding and early death. For the birth cohort 1960-1979, there is statistically significant (negative) association between inbreeding and early death. We have no explanation for the sudden change in the direction of the association, but it is quite strong and deserves further study. The accuracy of records for death dates may vary over time, so it may be dangerous to compare rates of early death between different cohorts. However, we have no reason to believe that the accuracy of the records varies widely within a 20-year cohort or that the accuracy varies depending on the inbreeding level. Therefore, we do not have concern that the associations we see are spuriously due to varying accuracy of records.

Winning. The relationship between being a mother of twins and inbreeding is summarized in Table 7. There is a slight trend towards more inbreeding in the controls (who are not mothers of twins), but this is not statistically significant. These results are consistent with those of Holloway and Sofaer (1990) on an outbred Scottish population, and the modeling of Meulemans et al. (1996) based on an Australian twin registry. Our results are not consistent with those of Philippe (1985) on a small sample from Quebec. The modeling of Meulemans suggested that a gene passed through the maternal line and acting dominantly may influence twinning. One would not expect to see much effect from such a gene in an inbred population, but if there were any effect it would make those mothers who are less inbred, and hence more likely to be heterozygous, more likely to have twins.

Discussion

We announce the creation of the Anabaptist Genealogy Database version 3.0 containing 295,095 individuals from three diverse original sources. Because much of the research interest in Anabaptist communities is due to their inbred nature, we chose to focus on evaluation of inbreeding and its effects to validate the enlarged data base. We used AGDB 3.0 to systematically and automatically reexamine previous analyses of inbreeding concerning: (1) consanguineous marriages over time, (2) family size, (3) birth intervals and reproductive compensation, (4) early death, and (5) twinning.

In this paper, we showed that over time there is a steady increase in both the fraction of consanguineous marriages and the amount of inbreeding in consanguineous marriages. We show positive association between (i) inbreeding and family size in the 20th century, (ii) inbreeding and interbirth intervals in the 20th century, and (iii) inbreeding and early death for individuals born between 1920 and 1959. We show negative association between inbreeding and early death for individuals born between 1960 and 1979 and found no association between inbreeding and being the mother of twins.

The associations we showed should not be interpreted as direct or causal, and the associations could be even stronger with any number of factors that are not fully recorded in the data sources. For example, there could be associations between inbreeding and any of the following: county of residence, Amish/Mennonite, subgroup within the Amish or Mennonites, and occupation. Khoury et al. (1987c, Table 4) compared inbreeding for farmers vs. nonfarmers, but the vast majority of the entries in their data base and ours had no occupation listed, so it is hard to draw any conclusions.

The increase in the geographical coverage and size of AGDB 3.0 compared to that of AGDB 2.0 make version 3.0 a more valuable data base to construct pedigrees automatically. AGDB 3.0 is also a useful tool to address demographic questions that are usually limited by the number of samples available, some of which we presented in this paper.

Acknowledgments Thanks to Mr. James Hostetler for providing the raw data file, which we processed to construct AGDB 3.0. Thanks to Dr. Judith Westman for putting us in contact with Mr. Hostetler. Thanks to Dr. Leslie Biesecker for collaborating on the design of AGDB and on several uses of it.

Received 21 November 2000; revision received 10 March 2001.

Literature Cited

Agarwala, R., L.G. Biesecker, K.A. Hopkins et al. 1998. Software for constructing and verifying pedigrees within large genealogies and an application to the Old Order Amish of Lancaster County. Genome Res. 8:211-221.

Agarwala, R., L.G. Biesecker, IF. Tomlin et al. 1999. Towards a complete North American Anabaptist genealogy: A systematic approach to merging partially overlapping genealogy resources. Am. J. Med. Genet. 86:156-161.

Al-Hendy, A., 0. Moshynska, A. Saxena et al. 2000. Association between mutations of the folliclestimulating-hormone receptor and repeated twinning. Lancet 356:914-914.

Becker, R.A., J.M. Chambers, and A.R. Wilks. 1988. The New S Language. Pacific Grove, CA: Wadsworth and Brooks/Cole.

Beiler, K. 1988. Fisher Family History. Lancaster, PA: Eby's Quality Printing.

Bortolus, R., F. Parazzini, L. Chatenoud et al. 1999. The epidemiology of multiple births. Human Reproduction Update 5:179-187.

Egeland, J.A. 1972. Descendants of Christian Fisher and Other Amish-Mennonite Pioneer Families. Baltimore, MD: Moore Clinic.

Gingerich, H.F., and R.W. Kreider. 1986. Amish and Amish Mennonite Genealogies. Gordonville, PA: Pequea Publishers.

Holloway, S.M., and J.A. Sofaer. 1990. Coefficients of relationship by isonymy among the parents of Scottish twins: A test of familial aggregation. Hum. Biol. 62:429-435.

Hsueh, W.C., B.D. Mitchell, J.L. Schneider et al. 2000a. QTL influencing blood pressure maps to the region of PPH1 on chromosome 2q31-34 in Old Order Amish. Circulation 101: 2810-2816.

Hsueh, W.C., MT Wagner, B.D. Mitchell et al. 2000b. Diabetes in the Old Order Amish: Characterization and heritability analysis of the Amish Family Diabetes Study. Diabetes Care 23:595-601.

Johnston, J.J., R.I. Kelley, T.O. Crawford et. al. 2000. A novel nemaline myopathy in the Amish caused by a mutation in troponin TI. Am. J. Hum. Genet. 67:814-821.

Khoury, MT, B.H. Cohen, G.A. Chase et al. 1987a. An epidemiologic approach to the evaluation of the effect of inbreeding on prereproductive mortality. Am. J. Epidemiol. 125:251-262. Khoury, MT, B.H. Cohen, E.L. Diamond et al. 1987b. Inbreeding and prereproductive mortality in

the Old Order Amish I. Genealogic epidemiology of inbreeding. Am. J. Epidemiol. 125:453-461.

Khoury, MT, B.H. Cohen, E.L. Diamond et al. 1987c. Inbreeding and prereproductive mortality in the Old Order Amish II. Genealogic epidemiology of prereproductive mortality. Am. J. Epidemiol. 125:462-472.

Khoury, M.J., B.H. Cohen, E.L. Diamond et al. 1987d. Inbreeding and prereproductive mortality in the Old Order Amish III. Direct and indirect effects of inbreeding. Am. J. Epidemiol. 125:473-483.

McKusick, V.A. 1978. Medical Genetic Studies of the Amish: Selected Papers. Baltimore, MD: Johns Hopkins University Press.

Meulemans, W.J., C.M. Lewis, D.I. Boomsma et al. 1996. Genetic modelling of dizygotic twinning in pedigrees of spontaneous dizygotic twins. Am. J. Med. Genet. 61:258-263.

Ober, C., T. Hyslop, and W.W. Hauck. 1999. Inbreeding effects on fertility in humans: Evidence for reproductive compensation. Am. J. Hum. Genet. 64:225-231.

Parisi, P., M. Gatti, G. Prinzi et al. 1983. Familial incidence of twinning. Nature 304:626-628. Philippe, P. 1985. Genetic epidemiology of twinning: A population-based study. Am. J. Med. Genet. 20:97-105.

Schull, W.V., and J.V. Neel. 1972. The effects of parental consanguinity and inbreeding in Hirado, Japan. V Summary and interpretation. Am. J. Hum. Genet. 24:425-453.

Stone, D., R. Agarwala, A.A. Schaffer et al. 1998. Genetic and physical mapping of the McKusickKaufman syndrome. Hum. Molec. Genet. 7:475-481.

Stone, D.L., A. Slavotinek, G.G. Bouffard et al. 2000. Mutation of a gene encoding a putative chaperonin causes McKusick-Kaufman syndrome. Nat. Genet. 25:79-82.

Weir, B.S. 1996. Genetic Data Analysis II. Sunderland, MA: Sinauer Associates.

RICHA AGARWALA,1 ALEJANDRO A. SCHAFFER,2 AND JAMES F. TOMLIN3

1Information Engineering Branch, National Center for Biotechnology Information, NIH, Bethesda, MD. 2Computational Biology Branch, National Center for Biotechnology Information, NIH, Bethesda, MD. 3Computational Bioscience and Engineering Laboratory, Center for Information Technology, NIH, Bethesda, MD.

Copyright Wayne State University Press Aug 2001
Provided by ProQuest Information and Learning Company. All rights Reserved