The Algebra of Mendelian Genetics
Cold Spring Harbor and the Forgotten Doctoral Equations of Population Inheritance
Bound inside the Massachusetts Institute of Technology library archives sits an unpublished 1940 doctoral dissertation titled An Algebra for Theoretical Genetics. After watching Claude Shannon turn nineteenth-century symbolic logic into a working calculus for telephone switches in 1937, his mentor Vannevar Bush urged him to try an even bolder jump across academic boundaries. Bush was appointed president of the Carnegie Institution of Washington in 1938 and formally took office in January 1939. In that post, Bush oversaw the Carnegie Department of Genetics at Cold Spring Harbor Laboratory on Long Island, New York. He asked Shannon whether the rules of Mendelian gene transmission could be captured by a special symbolic algebra just as cleanly as relay circuits obeyed George Boole's laws.
To learn how working biologists tracked hereditary traits, Shannon transferred from MIT's electrical engineering department into the mathematics department under algebraist Frank Lauren Hitchcock. He then traveled to Cold Spring Harbor for the summer of 1939. At Cold Spring Harbor, experimentalists such as maize cytogeneticist Barbara McClintock were mapping chromosome breakage, crossover frequencies, and ring chromosomes under the microscope. Shannon worked directly alongside psychologist and human geneticist Barbara Stoddard Burks. Burks was investigating whether complex human traits and hereditary anomalies showed chromosomal linkage across family pedigrees.
Classical population genetics in the 1930s, pioneered by Ronald Fisher, J. B. S. Haldane, and Sewall Wright, relied on cumbersome scalar recurrence equations. For a single genetic locus with two alleles $A$ and $a$, G. H. Hardy and Wilhelm Weinberg had proved in 1908 that random mating reaches stable genotype frequencies in a single generation. Once a geneticist tracked two or three linked gene loci situated along the same chromosome, however, chromosomal crossover during meiosis scrambled the allele combinations at a specific recombination fraction $r$. Predicting the exact distribution of genotypes ten generations into the future required pages of tedious element-by-element algebra.
In his 1940 dissertation, Shannon replaced those scalar recurrences with a non-associative, commutative genetic algebra:
- * Population vectors: Any diploid population distribution across multiple linked loci and multiple alleles is represented as a single multidimensional symbol $\Phi$ whose components sum to unity.
- * The Mendelian multiplication operator: Random mating between two parent populations $\Phi$ and $\Psi$ is written as the algebraic product $\Phi \times \Psi$, where the multiplication table directly encodes chromosomal crossover probabilities.
- * Recombination matrices: By decomposing the multi-locus chromosome state into linear combinations of single-locus marginal distributions and linkage disequilibrium residuals, each generation acts as a diagonalizable decay factor $(1 - r)^n$.
- * Three-locus closed-form forecasts: Using his symbolic notation, Shannon solved the exact $n$-generation trajectory for two-locus and three-locus linked traits with arbitrary numbers of alleles under random mating.
Shannon also examined lethal recessive genes, selection pressures, and sex-linked inheritance before submitting the dissertation to MIT in the spring of 1940. Because he soon pivoted to wartime defense research on anti-aircraft gun directors and cryptography, he never bothered to submit the chapters to a biological journal. Working independently at the University of Edinburgh between 1939 and 1941, mathematician I. M. H. Etherington introduced the formal study of baric and genetic algebras. Etherington had no idea that an American engineer had simultaneously carried the multi-locus recombination derivation much further. Although Shannon never published his 1940 doctoral dissertation and left biology for Bell Laboratories the following year, population geneticists who rediscovered the bound MIT typescript decades later found that his multi-locus recombination algebra had solved three-locus generational forecasts years ahead of the field.