Algebraic computation of genetic patterns related to three-dimensional evolution algebras

arXiv:1705.04157v1 [math.RA] 11 May 2017 Algebraic computation of genetic patterns related to three-dimensional evolution algebras1 O. J. Falc´on1, R. M. Falc´on2J. N´u˜nez3 1,3Faculty of Mathematics, University of Seville, Spain. 2School of Building Engineering, University of Seville, Spain. E-mail: [email protected], 2raf[email protected], 3jnvald[email protected] Abstract. The mitosis process of an eukaryotic cell can be represented by the structure constants of an evolution algebra. Any isotopism of the latter corresponds to a mutation of genotypes of the former. This paper uses Computational Algebraic Geometry to determine the distribution of three-dimensional evolution algebras over any field into isotopism classes and hence, to describe the spectrum of genetic patterns of three distinct genotypes during a mitosis process. Their distribution into isomorphism classes is also determined in case of dealing with algebras having a onedimensional annihilator. Keywords: Computational Algebraic Geometry, evolution algebra, classification, isotopism, isomorphism. 2000 MSC: 17D92, 68W30. 1 Introduction Evolution algebras were introduced by Tian and Vojtechovsky [22, 23] to simulate algebraically the self-reproduction of alleles in non-Mendelian Genetics. In the last years, these algebras have been widely studied without probabilistic restrictions on their structure constants [3, 6, 10, 8, 13, 17, 19, 18, 21]. A main problem to be solved in this regard is the distribution of evolution algebras into isomorphism classes. Besides, Holgate and Campos [7, 16] exposed the importance in Genetics of considering also the distribution of such algebras under isotopism classes, because they constitute a way to formulate algebraically the mutation of genotypes in the inheritance process. The mentioned distribution into isomorphism classes has already been dealt with for two-dimensional evolution algebras over the complex field [5, 9] 1Preliminary version. Accepted manuscript for publication in Applied Mathematics and Computation. DOI: 10.1016/j.amc.2017.05.045. 1 and for nilpotent evolution algebras of dimension up to four over arbitrary fields [15]. More recently, the authors [14] have characterized the isomorphism classes of two-dimensional evolution algebras over arbitrary fields and have established their distribution into isotopism classes. The latter gives rise to the spectrum of genetic patterns of two distinct genotypes during mitosis of eukaryotic cells. More specifically, if uand vare two distinct non-zero elements of the algebra under consideration, then the mentioned spectrum is formed by the next four genetic patterns: (0,0), for which no offspring exists; (u, 0), for which exactly one of the two genotypes does not produce offspring; (u, u), for which the offspring has always the same genotype, whatever the initial one is; and (u, v), for which the genotype of the offspring depends directly on that of the cell parent. This spectrum was obtained in [14] by means of distinct aspects on Computational Algebraic Geometry, all of them based on the fact that the algebraic law defined by the structure constants of any evolution algebra, together with the relations among basis vectors described by any isotopism of algebras, constitutes the set of generators of an ideal of polynomials whose reduced Gr¨obner basis establishes the algebraic relations that must hold, up to mutation, the genetic patterns of the mitosis process. This paper also focuses on the computation of such relations, particularly in case of dealing with three-dimensional evolution algebras. This enables us to distribute these algebras into isotopism classes, whatever the base field is, and describe mathematically the spectrum of genetic patterns of three distinct genotypes during a mitosis process. The distribution of such algebras into isomorphism classes is also determined in case of dealing with algebras having a one-dimensional annihilator. 2 Preliminaries In this section we expose some basic concepts and results on evolution algebras, isotopisms and Computational Algebraic Geometry that are used throughout the paper. For more details about these topics we refer the reader to the respective manuscripts of Tian [22], Albert [1] and Cox et al. [11]. 2.1 Evolution algebras Agene is the molecular unit of hereditary information. This consists of deoxyribonucleic acid (DNA), which contains the code to determine the attributes or phenotypes that characterize each organism. Genes are disposed in sequential order giving rise to long strands of DNA called chromosomes. 2 Genes related to a given phenotype can have distinct forms, which are called alleles. They always appear at the same position in chromosomes and constitute the genotype of the organism with respect to such a phenotype. In eukaryotic cells, DNA is primordially contained in the nucleus, although it also appears in organelles as mitochondria and chloroplasts, which are located in the cytoplasm. Mitosis is an asexual form of reproduction that consists of the division of an eukaryotic cell into two daughter cells in such a way that the nuclear genetic material of the former duplicates giving rise, up to rare mutation, to homologous nuclear chromosomes. In the final stage, the two daughter cells split having each resulting cell its corresponding copy of nuclear genetic material, whereas the extra-nuclear genetic material in the cytoplasm of the parent cell is randomly distributed between both of them by vegetative division. There exist distinct probabilistic laws that regulate the theoretical influence of all of this genetic material in the genotype of the offspring. Tian and Vojtechovsky [23] introduced evolution algebras to represent mathematically these laws. Specifically, an evolution algebra defined on a set β={e1,...,en}of distinct genotypes with respect to a given phenotype of an asexual organism is an n-dimensional algebra of basis βover a field Ksuch that eiej= 0, if i6=j, and eiei=Pn j=1 cijej, for some ci1,...,cin ∈K. The elements cij are called the structure constants of the algebra. Here, the product eiej= 0, for i6=j, is due to the uniqueness of the genotype of the parent cell; the product eieirepresents self-replication in the mitosis process; and each structure constant cij constitutes the probability, due to vegetative division, that the genotype eibecomes the genotype ejin the next generation. 2.2 Isotopisms of evolution algebras Two n-dimensional algebras Aand A′are isotopic [1] if there exist three nonsingular linear transformations f,gand hfrom Ato A′such that f(u)g(v) = h(uv), for all u, v ∈A. The triple (f, g, h) is called an isotopism between A and A′. If f=g, then the triple (f, f, h) is called a strong isotopism and the algebras are said to be strongly isotopic. If f=g=h, then the isotopism constitutes an isomorphism, which is denoted by finstead of (f, f, f). To be isotopic, strongly isotopic or isomorphic are equivalence relations among algebras. Throughout the paper, we refer the former and the latter as ∼ and ∼ =, respectively. Isotopisms of evolution algebras can be interpreted as mutations of the genetic material of parent and daughter cells in the mitosis process with 3 respect to a given phenotype. Specifically, if (f, g, h) is an isotopism between two n-dimensional evolution algebras Aand A′, then fand grepresent the respective possible mutation of each one of the two homologous chromosomes in which the nuclear genetic material of the parent cell duplicates during the first part of the mitosis process, whereas hrepresents a possible mutation of the genotype of the offspring in the final step of the process. For each α∈ {f, g, h}, the corresponding expression α(ei) = Pn j=1 aijejinvolves the genotype eito mutate to ejwith probability aij. Since A′is also an evolution algebra, the mitosis process only finishes if the genotypes of both homologous chromosomes that have been created after the mutations fand gcoincide. If these genotypes do not coincide, then there is no offspring. Hereafter, En(K) denotes the set of n-dimensional evolution algebras over the base field Kwith basis {e1,...,en}, whereas Tn(K) denotes the direct product Qn i=1he1,...,eni. Every evolution algebra in En(K) is uniquely determined by an structure tuple T= (t1,...,tn)∈ Tn(K), where ti=eiei, for all i≤n. The structure tuple Talso determines the genetic pattern of the corresponding mitosis process. Proposition 2.1 ([14]) Let Kbe a field. The next results hold. a) Any two structure tuples in Tn(K)that are equal up to permutation of their components and basis vectors give rise to a pair of strongly isotopic evolution algebras. b) Let Tbe a structure tuple in Tn(K). There always exists a structure tuple T′= (Pn j=1 c1jej,...,Pn j=1 cnjej)∈ Tn(K)such that a) If cii = 0, for some i≥1, then cjk = 0, for all j, k ≥i. b) If cii 6= 0, for some i≥1, then cij = 0, for all j6=i. c) The evolution algebra in En(K)of structure tuple T′is strongly isotopic to that one of structure tuple T. Let Abe an evolution algebra in En(K). Isotopisms preserve the dimension of the derived algebra A2={uv |u, v ∈A} ⊆ Aand that of the annihilator Ann(A) = {u∈A|uv = 0,for all v∈A}. Hereafter, En;m(K) denotes the subset of evolution algebras in En(K) with an m-codimensional annihilator. Theorem 2.2 ([14]) Let Kbe a field. The next results hold. a) Let m < m′. None evolution algebra in En;m(K)is isotopic to an evolution algebra in En;m′(K). 4 b) The set En;0(K)is only formed by the n-dimensional trivial algebra, which have all its structure constants equal to zero. c) Any algebra in En;1(K)is isotopic to the algebra described as e1e1=e1. If the former is not isomorphic to the latter, then it is isomorphic to the evolution algebra described as e1e1=e2. d) Any algebra in En;2(K)is isotopic to the algebra described as e1e1= e2e2=e1or to that described as e1e1=e1and e2e2=e2, and is isomorphic to an evolution algebra in En;2(K)such that e1e1∈ {e1, e2, e1+e2}. 2.3 Computational Algebraic Geometry Let Ibe an ideal of a multivariate polynomial ring K[X]. The algebraic set defined by Iis the set of common zeros of all its polynomials. The largest monomial of a polynomial in Iwith respect to a given monomial term ordering is its leading monomial. The reduced Gr¨obner basis of Iis the only subset Gof monic polynomials in Iwhose leading monomials generate the ideal also generated by all the leading monomials of Iand such that no monomial of a polynomial in Gis generated by the leading monomials of the rest of polynomials in the basis. This can always be computed from the sequential multivariate division of polynomials described by Buchberger’s algorithm [2], which is extremely sensitive to the number of variables [20]. Computational Algebraic Geometry can be used to determine the set of isotopisms and isomorphisms between two evolution algebras Aand A′in En(K), with respective structure constants cij and c′ij. To this end, we define the sets of variables Fn={fij |i, j ≤n},Gn={gij |i, j ≤n}and Hn= {hij |i, j ≤n}, which play the role of the entries in the regular matrices related to a possible isotopism (f, g, h) between Aand A′. The coefficients of each basis vector elin the expression f(ei)g(ej) = h(eiej) enable us to ensure that Pn k=1 fikgjkc′ kl is equal to 0, if i6=j, or to Pn k=1 hklcik, otherwise. Theorem 2.3 ([14]) The set of isotopisms and isomorphisms between A and A′are respectively identified with a) The subset of zeros (f11,...,fnn, g11,...,gnn, h11,...,hnn)∈K3n2in the algebraic set defined by the ideal IA,A′of K[Fn∪Gn∪Hn]described as h n X k=1 fikgjkc′ kl |i, j, l ≤n;i6=ji+h n X k=1 fikgikc′ kl − n X k=1 hklcik |i, l ≤ni, giving rise to non-singular matrices F= (fij),G= (gij )and H= (hij). 5 b) The subset of zeros (f11,...,fnn)∈Kn2in the algebraic set defined by the ideal JA,A′of K[Fn]described as h n X k=1 fikfjkc′ kl |i, j, l ≤n;i6=ji+h n X k=1 f2 ikc′ kl − n X k=1 fklcik |i, l ≤ni, giving rise to a non-singular matrix F= (fij). The computation of the reduced Gr¨obner bases of both ideals in Theorem 2.3 has recently been implemented [14] in the open computer algebra system for polynomial computations Singular [12]. Particularly, in case of dealing with the finite field K=Fq, with qa power prime, the respective complexity times that Buchberger’s algorithm requires to compute the bases under consideration are qO(3n2)+O(n8) and qO(n2)+O(n8). We have made use of the mentioned procedure to determine all the isotopisms and isomorphisms that appear throughout this paper. The run time and memory usage that are required to compute each one of the reduced Gr¨obner bases indicated in the paper are respectively 0 seconds and 0 MB in a computer system with an Intel Core i7-2600, with a 3.4 GHz processor and 16 GB of RAM, except for the computation related to Proposition 3.1.d, for which these two measures of computational efficiency have been 3 seconds and 1 MB, respectively. 3 Distribution of En;3(K)into isotopism classes Let n≥3 and let Kbe a field. In order to determine the distribution of the set En;3(K) into isotopism classes, Proposition 2.1.b enables us to focus on those n-dimensional evolution algebras described as Aa,b :=      e1e1=e1, e2e2=αe1, e3e3=βe1, for some α, β ∈K\ {0}, or Bα,β,γ :=      e1e1=e1, e2e2=e2, e3e3=αe1+βe2+γe3, for some (α, β, γ)∈K3\ {(0,0,0)}. 6 Proposition 3.1 The next results hold. a) Aα,β ∼A1,1, for all α, β ∈K\ {0}. b) Let α, β, γ ∈Kbe such that γ6= 0. Then, Bα,β,γ ∼B0,0,1. c) Let (α, β)∈K2\ {(0,0)}. Then, Bα,β,0∼B1,β′,0, for some β′∈K. d) B1,β,0∼B1,1,0, for all β∈K\ {0}. e) The evolution algebras B1,1,0and B1,0,0are not isotopic. Proof. Let us prove each assertion separately. a) The triple (f, Id,Id) such that f(e1) = e1,f(e2) = αe2and f(e3) = βe3 is an isotopism between the evolution algebras Aα,β and A1,1. b) Since isotopisms preserve the dimension of derived algebras, Proposition 2.1.b establishes that any evolution algebra Bα,β,γ with a threedimensional derived algebra (that is, such that γ6= 0) is isotopic to the evolution algebra B0,0,1. c) From Proposition 2.1.a, we can suppose α6= 0. Otherwise, it is enough to switch the vectors e1and e2. The triple (f, Id,Id) such that f(ei) = ei, for i∈ {1,2}, and f(e3) = αe3, is then an isotopism between the evolution algebras Bα,β,0and B1,β/α,0. d) The triple (f, g, h) related to the non-singular matrices F=  0β0 1 0 0 0 0 β G=  0 1 0 1 0 0 0 0 1  H=  0β0 100 001  is an isotopism between the evolution algebras B1,β,0and B1,1,0. e) Let (f, g, h) be an isotopism between the evolution algebras B1,1,0and B1,0,0and let H= (hij) be the non-singular matrix that is related to the linear transformation h. The computation of the reduced Gr¨obner basis related to the ideal IB1,1,0,B1,0,0in Theorem 2.3.a enables us to ensure in particular that h12 =h13 =h22 =h23 = 0. But then, the matrix His singular, which contradicts the fact of being (f, g, h) an isotopism. Hence, the algebras B1,1,0and B1,0,0are not isotopic.  Theorem 3.2 There exists four isotopism classes in En;3(K), whatever the base field is. They correspond to the evolution algebras A1,1,B0,0,1,B1,1,0 and B1,0,0. 7 Proof. The result follow straightforwardly from Proposition 3.1.  Corollary 3.3 There exist eight isotopism classes of three-dimensional evolution algebras over any field. They correspond to the evolution algebras of structure tuples (0,0,0),(e1,0,0),(e1, e1,0),(e1, e2,0),(e1, e1, e1),(e1, e2, e3), (e1, e2, e1+e2)and (e1, e2, e1). Proof. The result is an immediate consequence of Theorems 2.2 and 3.2.  Let us describe the spectrum of genetic patterns of three distinct genotypes during a mitosis process that are related to the classification exposed in Corollary 3.3. To this end, let u,vand wbe three distinct elements of the evolution algebra under consideration. The spectrum is formed by the next eight genetic patterns: (0,0,0), for which no offspring exists; (u, 0,0), for which only one of the genotypes gives rise to offspring; (u, u, 0), for which exactly one of the genotypes does not produce offspring, whereas the other two give rise to offspring with the same genotype; (u, v, 0), for which exactly one of the genotypes does not produce offspring, whereas the other two give rise to offspring with distinct genotypes; (u, u, u), for which the offspring has always the same genotype, whatever the initial one is; (u, v, w), for which the genotype of the offspring depends directly on that of the cell parent; (u, v, 1 2u+1 2v), for which the third genotype gives rise to each one of the genotypes produced by the other two with the same probability; and (u, v, u), for which two of the genotypes produce offspring with the same genotype. 4 Distribution of three-dimensional evolution algebras with a non-trivial annihilator into isomorphism classes Similarly to the reasoning exposed in the previous section, the distribution of the set E3(K) into isomorphism classes can be partitioned into that of the subsets E3;m(K), for m≤3. In particular, Theorem 2.2 describes these partitions in case of dealing with m∈ {0,1}. Specifically, the only evolution algebra in E3;0(K) is the trivial three-dimensional algebra, whereas there exist exactly two isomorphism classes in E3;1(K). They correspond to the three-dimensional algebra described by e1e1=e1and that one described by e1e1=e2. This section deals, therefore, with the computation of isomorphism classes in E3;2(K). To this end, Theorem 2.2.d enables us to focus on 8 those three-dimensional evolution algebras described as Cα,β,γ,δ,ǫ := (e1e1=αe1+βe2, e2e2=γe1+δe2+ǫe3, or Dα,β,γ,δ,ǫ := (e1e1=αe1+βe2, e3e3=γe1+δe2+ǫe3, for some (α, β)∈ {0,1}2\ {(0,0)}and (γ, δ, ǫ)∈K3\ {(0,0,0)}. The next result establishes when an evolution algebra of type Dα,β,γ,δ,ǫ is isomorphic to one of type Cα,β,γ,δ,ǫ. Throughout the distinct proofs of this and the subsequent results, we denote, respectively, as F= (fij) and G, the nonsingular matrix related to an isomorphism fbetween the algebras under consideration and the corresponding reduced Gr¨obner basis described in Theorem 2.3.b. Proposition 4.1 Let (α, β)∈ {0,1}2\{(0,0)}and (γ, δ, ǫ)∈K3\{(0,0,0)}. Then, a) D1,0,γ,δ,ǫ ∼ =C1,0,γ,ǫ,δ b) D1,1,γ,δ,ǫ ∼ =C1,0,γ,ǫ,δ−γ. c) If ǫ6= 0, then D0,1,γ,δ,ǫ ∼ =Cǫ,γ,0,0,1. d) If γ6= 0, then D0,1,γ,δ,0∼ =C0,γ,0,0,1. e) Let δ′∈K\ {0}. Then, i. The evolution algebras D0,1,0,δ′,0and Cα,β,γ,δ,ǫ are not isomorphic. ii. If δ6= 0, then the evolution algebras D0,1,0,δ,0and D0,1,0,δ′,0are isomorphic if and only if there exists m∈K\ {0}such that δ=m2δ′. Proof. Let us prove each assertion separately. a) It is enough to switch the basis vectors e2and e3. b) The result follows from (a) and the fact of being D1,1,γ,δ,ǫ ∼ =D1,0,γ,δ−γ,ǫ by means of the isomorphism fsuch that f(e1) = e1−e2and f(ei) = ei, for all i∈ {2,3}. c) It is enough to consider the isomorphism fsuch that f(e1) = e2,f(e2) = e3and f(e3) = e1−δe3/ǫ. 9 Acknowledgements The authors are very grateful to the Editors and to the referees for their valuable comments and suggestions that have improved the original version of the paper. Particularly, the authors gratefully thank to one of the referees for the suggestion of considering for our study the recent work developed by Cabrera et al. [4], which has enabled us to detect a pair of omitted cases of study in the preliminary version of the paper. References [1] A. A. Albert, Non-Associative Algebras: I. Fundamental Concepts and Isotopy, Ann. of Math., Second Series 43 (1942) 685–707. [2] B. Buchberger, An algorithm for finding the basis elements of the residue class ring of a zero dimensional polynomial ideal, J. Symbolic Comput. 41 (2006) 475–511. [3] Y. Cabrera Casado, M. Siles Molina, M. V. Velasco, Evolution algebras of arbitrary dimension and their decompositions, Linear Algebra App. 495 (2016) 122–162. [4] Y. Cabrera Casado, M. Siles Molina, M. V. Velasco, Description of three-dimensional evolution algebras. arXiv:1701.07219 (2017). [5] L. M. Camacho, J. R. G´omez, B. A. Omirov, R. M. Turdibaev, Some properties of evolution algebras, Bull. Korean Math. Soc. 50 (2013) 1481–1494. [6] L. M. Camacho, J. R. G´omez, B. A. Omirov, R. M. Turdibaev, The derivations of some evolution algebras, Linear Multilinear A. 61 (2013) 309–322. [7] T. M. M. Campos, P. Holgate, Algebraic Isotopy in Genetics, IMA J. Math. Appl. Med. Biol. 4(1987) 215–222. [8] J. M. Casas, M. Ladra, B. A. Omirov, U. A. Rozikov, On nilpotent index and dibaricity of evolution algebras, Linear Algebra Appl. 439 (2013) 90–05. [9] J. M. Casas, M. Ladra, B. A. Omirov, U. A. Rozikov, On Evolution Algebras, Algebra Colloq. 21 (2014) 331–342. [10] J. M. Casas, M. Ladra, U. A. Rozikov, A chain of evolution algebras, Linear Algebra Appl. 435 (2011) 852–870. 16 [11] D. A. Cox, J. B. Little, D. O’Shea, Using Algebraic Geometry, Springer- Verlag, New York, 1998. [12] W. Decker, G. M. Greuel, G. Pfister, H. Sch¨onemann, Singular 4-1-0 — A computer algebra system for polynomial computations, 2017. [13] A. Dzhumadildaev, B. A. Omirov, U. A. Rozikov, Constrained evolution algebras and dynamical systems of a bisexual population, Linear Algebra Appl. 496 (2016) 351–380. [14] O. J. Falc´on, R. M. Falc´on, J. N´u˜nez, Classification of asexual diploid organisms by means of strongly isotopic evolution algebras defined over any field, J. Algebra 472 (2017) 573–593. [15] A. S. Hegazi, H. Abdelwahab, Nilpotent evolution algebras over arbitrary fields, Linear Algebra App. 486 (2015) 345–360. [16] P. Holgate, Genetic algebras associated with polyploidy, Proc. Edinburgh Math. Soc. 15 (1966) 1–9. [17] A. Labra, M. Ladra, U. A. Rozikov, An evolution algebra in population genetics, Linear Algebra Appl. 457 (2014) 348–362. [18] M. Ladra, B. A. Omirov, U. A. Rozikov, Dibaric and evolution algebras in biology, Lobachevskii J. Math. 35 (2014) 198–210. [19] M. Ladra, U. A. Rozikov, Evolution algebra of a bisexual population, J. Algebra 378 (2013) 153–172. [20] Y. N. Lakshman, On the complexity of computing a Gr¨obner basis for the radical of a zero dimensional ideal. In: H. Ortiz (ed.) Proceedings of the twenty-second annual ACM Symposium on Theory of computing, STOC’90, New York, 1990; 555–563. [21] B. A. Omirov, U. A. Rozikov, K. M. Tulenbayev, On real chains of evolution algebras, Linear Multilinear Algebra 63 (2015) 586–600. [22] J. P. Tian, Evolution Algebras and their Applications, Lect. Notes Math. 1921, Springer-Verlag, Berlin, 2008. [23] J. P. Tian, P. Vojtechovsky, Mathematical concepts of evolution algebras in non-mendelian genetics, Quasigroups Related Systems 14 (2006) 111–122. 17