Bulletin of Mathematical Biology (2007) 69: 215–243 DOI 10.

1007/s11538-006-9119-3

ORIGINAL ARTICLE

An Extended RNA Code and its Relationship to the Standard Genetic Code: An Algebraic and Geometrical Approach
´ a,∗ , Eberto R. Morgadob , Tzipe Govezenskya Marco V. Jose
a

Theoretical Biology Group, Instituto de Investigaciones Biom´ edicas, Universidad Nacional Autonoma de M´ exico, M´ exico D.F. 04510, M´ exico ´ b Facultad de Matem´ atica, F´ ısica y Computacion, ´ Universidad Central “Marta Abreu” de Las Villas, Santa Clara, Cuba
Received: 19 September 2005 / Accepted: 23 February 2006 / Published online: 2 November 2006 C Society for Mathematical Biology 2006

Abstract An algebraic and geometrical approach is used to describe the primaeval RNA code and a proposed Extended RNA code. The former consists of all codons of the type RNY, where R means purines, Y pyrimidines, and N any of them. The latter comprises the 16 codons of the type RNY plus codons obtained by considering the RNA code but in the second (NYR type), and the third, (YRN type) reading frames. In each of these reading frames, there are 16 triplets that altogether complete a set of 48 triplets, which specify 17 out of the 20 amino acids, including AUG, the start codon, and the three known stop codons. The other 16 codons, do not pertain to the Extended RNA code and, constitute the union of the triplets YYY and RRR that we define as the RNA-less code. The codons in each of the three subsets of the Extended RNA code are represented by a fourdimensional hypercube and the set of codons of the RNA-less code is portrayed as a four-dimensional hyperprism. Remarkably, the union of these four symmetrical pairwise disjoint sets comprises precisely the already known six-dimensional hypercube of the Standard Genetic Code (SGC) of 64 triplets. These results suggest a plausible evolutionary path from which the primaeval RNA code could have originated the SGC, via the Extended RNA code plus the RNA-less code. We argue that the life forms that probably obeyed the Extended RNA code were intermediate between the ribo-organisms of the RNA World and the last common ancestor (LCA) of the Prokaryotes, Archaea, and Eucarya, that is, the cenancestor. A general encoding function, E, which maps each codon to its corresponding amino acid or the stop signal is also derived. In 45 out of the 64 cases, this function takes the form of a linear transformation F , which projects the whole six-dimensional hypercube onto a four-dimensional hyperface conformed by all triplets that end in cytosine. In the remaining 19 cases the function E adopts the form of an affine
∗ Corresponding author. ´ E-mail address: marcojose@biomedicas.unam.mx (M. V. Jose).

216

Bulletin of Mathematical Biology (2007) 69: 215–243

transformation, i.e., the composition of F with a particular translation. Graphical representations of the four local encoding functions and E, are illustrated and discussed. For every amino acid and for the stop signal, a single triplet, among those that specify it, is selected as a canonical representative. From this mapping a graphical representation of the 20 amino acids and the stop signal is also derived. We conclude that the general encoding function E represents the SGC itself. Keywords Primaeval RNA code · Standard genetic code · Evolution of the genetic code · Extended RNA codes · Algebra and geometry

1. Introduction The current genetic code is considered to be nearly universal. This code is written in an alphabet of four letters (C, A, U, G), grouped into words three letters long, called triplets or codons. Each of the 64 codons specifies one of the 20 amino acids or else serves as a punctuation mark signaling the end of a message. Given 64 codons and 20 amino acids plus a punctuation mark there are 2164 ≈ 4 × 1084 possible genetic codes. Is there something special about the only one code that governs all life on Earth? Francis Crick (1968) argued that the Standard Genetic Code (SGC) need not be special at all; it could be nothing more than a “frozen accident.” This concept is not far away from the idea that sometime there was an age of miracles. However, when the SGC was compared to a computer generated random sample of one million alternatives, the natural code emerged as superior to every random permutation with a single exception (Freeland and Hurst, 1998). Recently, numerical experiments with hand-crafted genetic codes analyzed in silico showed inferior statistical properties (such as information content, scaling and autocorrelation properties) than the SGC (Garc´ ıa et al., 2004). It is widely accepted that there was an age in the origin of life in which RNA played the role of both genetic material and main agent of catalytic activity (e.g. Woese, 1967; Crick, 1968; Kenneth and Ellington, 1995). This period is known as the RNA World (Gilbert, 1986; Gesteland et al., 1999). Investigations on the minimal gene set that is necessary and sufficient to sustain the existence of cellular life are consistent with the notion that the last common ancestor (LCA) of the three primary kingdoms (Archaea, Eucarya, and Prokaryotes) had an RNA genome (Mushegian and Koonin, 1996; Hutchinson et al., 1999; Gil et al., 2002). However, the quasi-species concept of Eigen and Schuster (1977) demonstrated that the accuracy of replication placed limits on the size of the genome that can be maintained by selection. The higher the error rate during replication, the smaller the maximum possible permissible genome size. Thus, replication fidelity was a strong limiting feature in the RNA World. On the other hand, sequence similarities shared by many ancient, large proteins found in all three kingdoms of life suggest that considerable fidelity already existed in the operative genetic system of their LCA, but such fidelity is unlikely, given the Eigen’s limit, to be found in RNA-based genetic systems (Lazcano, 1995; Lazcano and Miller, 1996). The cenancestor probably had a DNA genome (Becerra et al., 1997).

there has not been systematic studies that relate the RNA code (Crick. Fig. 2004a. 01. we used the following ordered assignment of the nucleotide bases: C↔ 00. (C. With this ordering.b. the constraint of having an intact message in only one reading frame has to be relaxed. A3. U. C.1. A. A. and an encoding function for the RNA-less code. We also derive the general encoding function of the SGC. (1957) but rather an RNA code which can be translated in the first. Interestingly. G). U). The main question that we address in this work is to see if via our algebraic and geometrical approach we can shed some light on the problem of how the primaeval RNA code could have evolved to generate the SGC.. and 01 is to 10.. 1996. 2005). two algebras were presented to reflect the relationship between codon assigment and the physicochemical aspects of the amino acids. U. about the concepts of algebra and geometry is provided at the end of this article. First. given that translational and transcriptional errors were probably of great importance early in the history of life. Therefore. G. U. the three four-dimensional hypercubes and the right hyperprism (A9. 3. whereas the pyrimidines C and U are represented by the even numbers 0 and 2. U↔ 10. 1968. 1973. Jimenez-Monta no ´ et al. A. U). Eigen and Schuster. Hence. and we show that this function is an integration of the different encoding functions of the above-mentioned four sets. associated to the Extended RNA code. G).1. observing that 64 is equal not only to 43 but also to 26 .Bulletin of Mathematical Biology (2007) 69: 215–243 217 To our knowledge. A and U are complementary to each other in double stranded DNA. second and third reading frames. Next. and.. we show that each reading frame can be represented as a four-dimensional hypercube (A3. it is convenient to select the ordering in such a way that 00 is complementary of 11. 1977. purines A and G are represented by the odd numbers 1 and 3. Several authors (e. G↔ 11. 11 to the letters C. 1. This encoding function adopts different forms for different subsets of codons. suggested to organize the codon table as a six-dimensional hypercube ´ or six-dimensional vector space over the binary field. 10. To this end. . G. C. Finally. (A. (A. 5). Here. A↔ 01. 2005). we search for symmetries and patterns in both the SGC and the RNA code.g. we consider not a strict comma-less code as proposed by Crick et al. may be done in 24 = 4! ways. we define an encoding function for each of the three four-dimensional hypercubes. For the interested reader. Theoretical background The standard table of codon assignments derives from the obvious representa´ ˜ tion of the triplet code as a 4 × 4 × 4 cube.1) as derived by concepts of combinatorial geometry.1. when the machinery of protein synthesis was imprecise. Sanchez et al. and it represents the SGC code itself.. G. This assignment of the duplets 00. we discuss our findings in terms of the origin and evolution of the SGC. The article is organized as follows. We hypothesize that in order to allow further evolution of the RNA genetic code.3) associated to the RNA-less code can be inserted as pairwise disjoint four-dimensional affine subspaces in the six-dimensional hypercube (Coxeter. which also represent the integers 0. Konecny et al. Given that C and G. 1995) with the SGC. in the binary numerical system. 2. 1. an Appendix which is referred to throughout this work. In previous works (Sanchez et al. we have only eight possible selections: (C. A1.

This sum. The 64 sextuples of zeros and ones generate the so-called six-dimensional hypercube. Results 2. C. This extended code includes 48 triplets which specify 17 amino Table 2 Sum module 2 of nucleotide bases + C A U G C C A U G A A C G U U U G C A G G U A C . we have selected the first. when they are portrayed in a plane. Y . G. This is so because the picture is a projection of a six-dimensional figure over a plane. all the angles between adjacent edges are supposed to be right angles.218 Table 1 + 0 1 Bulletin of Mathematical Biology (2007) 69: 215–243 Sum module 2 in the field Z2 0 0 1 1 1 0 (U. according to the terminology of Coxeter (1973) (A3). 2. 1}(A3. also called XOR operation. (U. A). U.1). widely used in symbolic logic. When Table 1 is translated to the nucleotide bases it gives Table 2. N . C. C). and (G. A). A. (G. The RNA code and its three reading frames: The extended code A primaeval RNA World was proposed (Gilbert. In the hypercube.purine. for the elements of the field Z2 . In the vector space structure of the hypercube the addition is the so-called sum module 2. but it is possible to show that the remaining ones lead to the same results. as in every orthotope. which is the operation table of a known abelian group of order 4. The same happens with some of the angles of a three-dimensional cube. but in our pictures many of them look like acute or obtuse. and codons with an NYR and YRN patterns emerged. The vector space (Z2 )6 is an orthotope. These three patterns altogether are here defined as the codons of the Extended RNA code. Since the primitive translational apparatus may have been imperfect.pyrimidine. 1986) which comprises the codons with an RNY pattern (R . A. which is a vector space over the binary field Z2 = {0.1. the so-called Klein Four Group. From these orderings. bit by bit. U. G. shifts in the reading frame probably occurred. is defined in Table 1. This group is isomorphic (A3) to the group of all the symmetries of a plane rectangle (not including the square).any nucleotide). C).

. and Z belong to the set {C. associated to the canonical vector e4 . the second (NYR pattern) and the third (YRN pattern) sets correspond to readings of the primaeval RNA code at the second and the third reading frames. A. They constitute coordinated lines or edges of the hypercube. the start codon. where X. 2. Eigen and Schuster. k). we denote as ti j the composed function ti ◦ t j . Now we will consider the combinatorial geometry. and we defined it as the RNA-less code or complementary code of the Extended RNA code. The set of codons of the form RNY is the union of the cubes RYY (Fig. we may consider two subclasses. 1978). For every triplet (i . The first set corresponds to the so-called primaeval RNA code (RNY pattern). U. associated to the structure of the vector space. we introduce the following notational convention. but also the same first nucleotide. the vector e2 + e4 .1). 1a) and RRY (Fig. and that is the subset YYY. In this context. Y. In most cases. the 64 triplets XYZ of the SGC. of the canonical base. that is. RRY is obtained from YYY by the addition of the triplet AAC. For each of the eight subsets. these subclasses correspond to coordinated planes or faces of the cube. The fourth set is the union of two subsets: YYY and RRR. these pairs of codons specify the same amino acid. For example. we denote as ti the associated translation: v → v + ei .1). and the fourth set corresponds to the RNA-less code (YYY and RRR patterns). First reading frame The 16 triplets of the RNY pattern code for eight amino acids and are considered as the primaeval RNY code (Crick. The eight elements of a subset correspond to a three-dimensional coordinated affine subspace (A4. and the three known stop codons.Bulletin of Mathematical Biology (2007) 69: 215–243 219 acids. We say that the reading and translation of sequences of codons of the form RNY defines the first reading frame (FRF) in the RNA World. j . for the set RNY we obtain the two subsets RYY and RRY. every pair consisting of triplets that have. we denote as ti jk the composed function ti ◦ t j ◦ tk and so on (A9. including AUG.2) in the first nucleotide. an ordinary cube.1) or vectorial cube. The other seven subsets are three-dimensional affine subspaces obtained from YYY by translations (A9. Each of the first three sets can be partitioned into two subsets by replacing the N by Y or R. the addition of the vector e2 which involves a transversion (A10. contained in the six-dimensional hypercube. not only the same central nucleotide. Each cube is obtained from the other by means of the translation t4 . each of them formed by triplets having the same central nucleotide. each of the subclasses is partitioned into two pairs. G}.1). RYY is obtained from YYY by the addition of the triplet ACC. 1968. For every pair (i . which represents transversions in the first and the second nucleotides. respectively. only one is a vector subspace (A1. that is. The remaining 16 codons can only be obtained by mutations other than by frame-shift readings. From the eight subsets. that is. 2004.2. j ). Finally. A6. which contains the null vector CCC . For any vector space the associated geometry is defined as the family of all the affine subspaces or linear varieties of the given space (A2). For every vector ei . In order to describe algebraically and geometrically all sets. can be partitioned into four sets of 16 triplets each. This is so since the Hamming distance (He et al. that . 1b) in which every amino acid corresponds to an edge of the associated cube.1) between triplets on the same edge is 1.

due to an slippage. ignoring the first nucleotide R. 1973).220 Bulletin of Mathematical Biology (2007) 69: 215–243 Fig. also called a measure polytope. The 16 triplets of the form NYR specify eight amino acids. the reading starts at the second nucleotide N. Graphical representation of the subsets (a) RYY and (b) RRY. the triplets of the form NYR give . is. and is therefore an orthotope (Coxeter. five of which were also found in the FRF. (c) First Reading Frame: RNY = RYYURRY. The second reading frame If in a sequence of triplets of the form RNY. It is a symmetrical regular polytope with mutually orthogonal (A8) sides. Note that in each cube the four amino acids correspond to 4 (solid edges) out of the 12 edges of the cube.3. 1c). or a four-dimensional hyperface of the sixdimensional hypercube of the 64 triplets (Fig. The set of triplets of the form NYR is a disjoint set with the set RNY. is equal to 1. 2. then triplets of the form NYR will be translated. which is a coordinated affine subspace (A4). Then. The hypercube is a generalization of a three-dimensional cube in n dimensions. 1 RNY code. the addition of the codon CAC to each triplet. It means that the set of codons of the form RNY constitutes a four-dimensional hypercube. The Hamming distance between a vertex and its image under the translation t4 .

that is. In this set. 2 NYR code. the addition of the vector e2 + e6 . rather there are two amino acids (Met and Ile) which correspond to single vertexes. 2a) and RYR (Fig. two amino acids correspond to 2 (solid edges) out of the 12 edges of the cube. that is. the start codon AUG ensues (Konecny et al. Note that in the cube YYR the amino acid Leu corresponds to a face and in the cube RYR not every amino acid corresponds to an edge of the associated cube. 2b). RYR is obtained from YYY by the addition of the triplet ACA. Note that in (a). those of the FRF and the SRF. and Leu correspond to a face (four connected solid edges). In (b) not all the amino acids correspond to edges of the cube. of the vector e6 which represents a transversion in the third nucleotide. The set of codons of the form NYR is the union of the cubes YYR (Fig. We consider the union of the sets of codons of the form RNY and NYR. YYR is obtained from YYY by addition of the codon CCA. and their correspondence with their 11 amino acids. as an extension of the original RNY code. that is. rise to only three new amino acids. which involves transversions in the first and third nucleotides. 1995).Bulletin of Mathematical Biology (2007) 69: 215–243 221 a) b) c) Fig. (c) Second Reading Frame: NYR = YYRURYR. Graphical representation of the subsets (a) YYR and (b) RYR. We say that the reading and translation of sequences of codons of the form NYR defines the second reading frame (SRF) in the RNA World. Each cube can be derived ..

without repetitions of any of those found in the FRF or the SRF. 2. the addition of the vector e4 + e6 which involves transversions in the second and third nucleotides. or double rotation. We will consider the union of the sets of codons of the form RNY. ignoring the first (R) and the second (N) nucleotides. Each cube is obtained from the other by the translation t6 . and the other three codons correspond to a termination signal. those associated to the first. of the vector e4 that represents a transversion in the second nucleotide. Then. which is a coordinated affine subspace. e2 → e6 . The set of codons of the form YRN is the union of the cubes YRY (Fig. YRY is obtained from YYY by addition of the codon CAC. e4 → e2 . where 45 out of them specify 17 of the primary amino acids. NYR and YRN. comprises the Extended RNA code. The hypercube NYR is the image of the RNY under the non-singular linear transformation e1 → e5 .4. It means that the set of codons of the form NYR constitutes also a four-dimensional hypercube. that is. associated to the canonical vector e2 . or four-dimensional hyperface of the six-dimensional hypercube of the 64 triplets (Fig. is equal to 1. YRR is obtained from YYY by the addition of the triplet CAA. The matrix of this linear transformation is orthogonal with determinant 1. In summary. e3 → e1 . or a four-dimensional hyperface of the six-dimensional hypercube of the 64 triplets (Fig. e6 → e4 . The set of triplets of the form YRN is a disjoint set with the sets RNY and NYR. The other three codons are the so-called stop codons. under the translation t6 . . The Hamming distance between a vertex and its image. second and third reading frames. e6 → e4 . by the addition of the codon ACC to every triplet. e2 → e6 . The Hamming distance between a vertex and its image. We say that the reading and translation of sequences of codons of the form YRN defines the third reading frame (TRF) in the RNA World. is equal to 1. NYR and YRN. It means that the set of codons of the form YRN constitutes also a four-dimensional hypercube. e5 → e3 . e3 → e1 . e4 → e2 . and their correspondence with 17 amino acids and stop codons. 2c). under the translation t2 . that is. which complete a set of 17 out of the 20 primary amino acids. In the YRY cube every amino acid corresponds to an edge of the associated cube but this is not the case in the YRR cube. 3a) and YRR (Fig. 3c). associated to the canonical vector e6 . the reading starts at the third nucleotide Y. The third reading frame If in a sequence of triplets of the form RNY. which is a coordinated affine subspace. which is defined by an even permutation of the canonical base. the triplets of the form YRN give rise to six new amino acids. which is the same basis permutation. which converts RNY into NYR. the addition of the codon CCA to every triplet.222 Bulletin of Mathematical Biology (2007) 69: 215–243 from the other by means of the translation t2 . and it can also be interpreted as a double rotation in the six-dimensional Euclidean vector space R6 (A7). as an extension of the primaeval RNY code. The hypercube YRN is the image of the NYR under the non-singular linear transformation e1 → e5 . the set of 48 codons of the form RNY. that is. This function maps every triplet XYZ onto the triplet YZX. due to a slippage. that is. 3b). e5 → e3 . Thirteen out of the 16 triplets of the form YRN code for six new amino acids. then triplets of the form YRN will be translated. that is.

Trp corresponds to only one vertex. these triplets code for eight amino acids. and the three stop codons belong to this cube connected by two solid edges. If we consider the reading and translation of sequences of codons of the form YYY or RRR. but they do not enter into the composition of the Extended RNA code. The codons of the RNA-less World The remaining 16 triplets belong to the cubes YYY or RRR. Then. The 16 codons . For this reason. The union of the Extended RNA code and its complementary RNA-less code. those composed by only pyrimidines or only purines. we will call them the triplets of the RNA-less World.5. five out of which are repetitions of those found in the Extended RNA code. the triplets of the form YYY or RRR give rise to only three new amino acids. 2. that is. four amino acids correspond to 4 (solid edges) out of the 12 edges of the cube. constitutes the six-dimensional hypercube of the 64 triplets with the 20 amino acids and a termination mark. 3 YRN code. This addition completes the set of the 20 amino acids. Graphical representation of the subsets (a) YRY and (b) YRR. In (b) two amino acids correspond to an edge of the cube (Arg and Gln).Bulletin of Mathematical Biology (2007) 69: 215–243 223 a) b) c) Fig. Note that in (a). (c) Third Reading Frame: YRN = YRYUYRR.

4b).224 Bulletin of Mathematical Biology (2007) 69: 215–243 a) b) c) Fig. that is. 4a) and RRR (Fig. the Hamming distance between a vertex and its . 4 The RNA-less code: YYYURRR. the four amino acids correspond also to 4 (solid edges) out of the 12 edges of the cube. Recall that YYY is a vector subspace of the whole vector space. the addition of the codon AAA to every triplet. (c) RNA-less code. In contrast to the hypercubes of the Extended RNA code. which performs a transversion in every nucleotide component. in both cubes (a) and (b). Graphical representation of the subsets (a) YYY and (b) RRR. of the RNA-less code represent the union of the cubes YYY (Fig. Each cube could be obtained from the other by the translation t246 associated to the vector e2 + e4 + e6 . Note that.

5 A graph representation diagram of the Boole lattice of the 64 triplets of the SGC.. purple • YRN (RNA code in the TRF). the icosahedron or dodecdimensional hypercube (Jimenez-Monta no ahedron (White undated). The sixdimensional hypercube can be envisaged as composed by: yellow • RNY (primaeval RNA code).g. a four-dimensional hyperface but rather it is a right hyperprism of height 3 (Fig. it has been customarily to represent in various graphical ways (e. the result of the union of the pairwise disjoint sets of the Extended RNA code plus the RNA-less code is the six-dimensional hypercube of the 64 codons of the SGC (Fig. Interestingly. 5).3). none of these Fig. Actually. It means that the set of codons of the RNA-less code do not lead to a four-dimensional hypercube. under the translation t246 . but it is not as in the other cases. orange • NYR (RNA code in the SRF). the six´ ˜ et al. 3. black • YYY and RRR (RNA-less code). or the standard table of the genetic code) the 64 codons of the SGC but without any reference to an encoding function that maps each triplet with its corresponding amino acid or stop signal. is equal to 3.Bulletin of Mathematical Biology (2007) 69: 215–243 225 image. Encoding functions So far. The set is a fourdimensional subspace of the six-dimensional hypercube. 4c and A9. . 1996).

that is. UAG. F ◦ F = F . AGU. CAG UAA. we derive encoding functions for the Extended RNA code. GUG AAC. CUG. GCA. t .CCA. the RNA-less code. the representative of Ala is the triplet GCC. UGU CAA. CUA. u. in fact. GGG CCC. CCG CUC. F : (x . and the SGC. v ). a vectorial four-dimensional hypercube or a hyperface of the six-dimensional hypercube. 0. 0.GGU. that is. AGA. that is. GUA. UCA. The image of F . it is well known that most mutations in the third base does not change the corresponding amino acid (the wobble hypothesis) and therefore we started with a function that maps the third nucleotide of a codon to zero. The endomorphism F belongs to a class of endomorphisms. GUU. We select the representation of every amino acid by only one triplet. UCC. For example. AGC UAC. GAG UUC. v ) → (x . ACU. The kernel of F is the two-dimensional subspace of vectors of the form (0. CCU. is the set of triplets of the form XYC.1. y. those that end with the nucleotide C. CGU. GCG GUC. Biologically. the kernel of F is a face of the hypercube. AAG GAA. ACG AUC. It is the solution subspace of the homogeneous linear . AAU AGC. the correspondence of the ordered set of triplets that specify every amino acid is illustrated. we consider the linear transformation. 0. 0). It is a four-dimensional subspace. denoted by Im( F ). An auxiliary linear function for the different encoding functions In Table 3. U. t . AUU GCC. The ordering of the set of triplets is the linear order determined by the selected order of the bases (C. 0. the first. ACA. CUU.226 Bulletin of Mathematical Biology (2007) 69: 215–243 Table 3 The correspondence between the ordered set of triplets and every amino acid and stop codons Amino acid Threonine RF Isoleucine Alanine Valine 1st Asparagine Serine Aspartic acid Glycine Proline RF Leucine 2nd Methionine Histidine RF Arginine Tyrosine 3rd Cysteine Glutamine Stop Tryptophan Lysine RNA Glutamic acid Phenylalanine less Cube Symbol Sextuple Set of triplets 1 1 1 1 2 2 2 2 3 3 4 5 5 5 5 6 6 6 7 7 8 Thr Ile Ala Val Asn Ser Asp Gly Pro Leu Met His Arg Tyr Cys Gln Stop Trp Lys Glu Phe 010000 011000 110000 111000 010100 011100 110100 111100 000000 001000 011011 000100 001100 100100 101100 000101 100101 101111 010101 110101 101000 ACC. Herein. z. UGA UGG AAA. the triplets of the form CCZ. Hence.UCG GAC.UUA. UUU representations is the genetic code. which are called projections and are characterized by the property of idempotency. CGA. which carries over every triplet XYZ to the triplet XYC. it changes the third nucleotide by the nucleotide C. 3. AUA. CAU CGC. of the vector space (Z2 )6 . For the three stop codons we select as representative the triplet UAA. CGG. or endomorphism F (A9. UCU.2). GAU GGC. UAU UGC. A. that is. GCU. GGA. y. z. u. UUG AUG CAC. G) so that the triplet at the left-most position is the one with the minimum value. Consequently.

is selected. image of the function F1 in the set RNY. We denote this hypercube as NNC (Fig. is the intersection of the four-dimensional hypercubes RNY . if compared with the other representative.2. or else while preserving the first and second nucleotides. the function F1 assigns to each of the 16 triplets of the set RNY the same triplet if it ends with C. e4 . This means that for every edge associated to the same amino acid. we define a function F1 . This set is a four-dimensional coordinated vector subspace. The function F1 is the restriction. G}. system u = 0. image of the function F . 6 The hypercube NNC . 3. This representative triplet in the cube RNC . e2 . t . is the minor. in fact. v which are the components of a generic vector in (Z2 )6 . y. which projects it onto its subspace of triplets of the form NNC. representing the FRF of the primaeval RNA code. a four-dimensional hyperface of the whole six-dimensional hypercube generated by the canonical vectors e1 . Note that the blue three-dimensional cube is image of F1 . v = 0. according to the linear order of the set of triplets as derived from the selected ordering in the set {C. Hypercube NNC image of the function F .Bulletin of Mathematical Biology (2007) 69: 215–243 227 Fig. z. e3 . its vertex that ends with the nucleotide C. which projects the whole hypercube onto the cube RNC . U. In other words. A. to the hypercube RNY. of the whole six-dimensional hypercube. The encoding function for the primaeval RNY code In the four-dimensional hypercube of codons of the form RNY. u. with unknowns x . of the linear transformation F : XYZ → XYC . the third one is changed to C if it is U. 6). Note that the cube RNC . The triplet that is selected as the canonical representative of each amino acid is the one which belongs to the cube RNC .

belong to the cube NYA. Here. in the following way. we define a function F3 . AUU → AUC GCC. UCG → UC A CC A. AAU → AAC AGC. GCU → GCC GUC. as is the case of F1 in the hypercube RNY. The function F2 is related to the linear transformation F . 3. But the function F2 is not exactly a linear projection of NYR onto NYA.4. whereas for AUG it acts as the composition t56 ◦ F . GCG → GC A AC A. with only three exceptions: UUG. described above. which assigns to every set of triplets. Actually. for 13 out of the 16 triplets. associated to the TRF of the Extended RNA code. we are taking the cube RNC as a canonical representation of the set of eight amino acids. according to the linear order in the whole set of triplets. In most cases. UUA and AUG. the composition of F with the translation t6 . CCG → CC A GU A. respectively. UUG → CU A UC A. CUG. F2 behaves as the composition t16 ◦ F . GUG → GU A AUG → AUG AU A → AU A GC A. F2 coincides with the restriction to NYR of the affine transformation t6 ◦ F . the minor. For UUG and UUA. specified by the F RF . ACG → AC A (for Leucine) (for Serine) (for Proline) (for Valine) (for Methionine) (for Isoleucine) (for Alanine) (for Threonine) We note that all of the images of the function F2 . which assigns to every set of triplets. ACU → ACC AUC. we define a function F2 . GAU → GAC GGC. which is a three-dimensional hyperface of the hypercube NYR.3. encoding for the same amino acid. AGU → AGC GAC. it is so. UU A. associated to the SRF of the Extended RNA code. GUU → GUC AAC. . The explicit definition of the function F1 is: F1 ACC. The encoding function for the triplets of the second reading frame In the hypercube of codons of the form NYR. with the only exception of AUG. GGU → GGC (for Threonine) (for Isoleucine) (for Alanine) (for Valine) (for Asparagine) (for Serine) (for Aspartic acid) (for Glycine) 3. The encoding function for the triplets of the third reading frame In the hypercube of codons of the form YRN.228 Bulletin of Mathematical Biology (2007) 69: 215–243 and NNC . The explicit definition of the function F2 is: F2 CU A. t16 and t56 the composed translations t1 ◦ t6 and t5 ◦ t6 .

F3 . CCC. CAG and CAA. F4 acts as the restriction to RRR of the affine transformation t6 ◦ F . UGA → U AA C AA. for the triplets AAG. However. The encoding function for the triplets of the RNA-less code In the vector subspace of the codons of the form RRR or YYY. CUU → CUC UCC. The linearity of the encoding functions is apparent even considering few departures. GAA. UCC. the function F3 coincides with the restriction to YRN of the linear transformation F : XYZ → XYC . the function F4 acts as the restriction to the set YYY of the linear transformation F . However. AAA. In order to build this plot. AGG → AGA GAA. CGG → CGC U AC. C AU → C AC UGG → UGG U AA. note that several codons have only . UAA. and F4 is shown in Fig. A graphical representation of the encoding functions F1 . F3 coincides with the restriction to YRN of the affine transformation t6 ◦ F For UGA. CUC. AAG → AAA AGA. 3. F2 . which assigns to every set of triplets. In fact. GAG. the minor. C AG → C AA (for Cysteine) (for Arginine) (for Tyrosine) (for Histidine) (for Tryptophan) (for Stop codons) (for Glutamic acid) Note that four out of the seven images for the function F3 . the minor. AGG. UUU → UUC AAA. CGA. UUC. The abscissa represents each of the 64 codons and they are mapped according to the encoding functions of each subset which results in their respective representative codons. CGU. CCU → CCC CUC. U AG. UCU. associated to the RNY-less code we can define a function F4 . GGG → GGA (for Proline) (for Leucine) (for Serine) (for Phenylalanine) (for Lysine) (for Arginine) (for Glutamic acid) (for Glycine) For the triplets CCU. GGG. we have considered the lexicographic order of the triplets used above and we have assigned to this order the corresponding integer values from 0 to 63. The explicit definition of the function F4 is: F4 CCC. 7. GGA. UGU → UGC CGC. For UGG. F3 coincides with the restriction to YRN of the affine transformation t36 ◦ F . belong to the cube YRC . UCU → UCC UUC. encoding for the same amino acid. for 10 triplets. For UAG. CUU. according to the linear order in the whole set of triplets. GAG → GAA GGA. The explicit definition of the function F3 is: F3 UGC. U AU → U AC C AC.5. UUU. according to the linear order in the whole set of triplets. F3 coincides with the restriction to YRN of the affine transformation t56 ◦ F .Bulletin of Mathematical Biology (2007) 69: 215–243 229 encoding for the same amino acid. AGA.

F2 and F4 . F3 and F4 . U ↔ 10. Herein. hence these codons have two or three representative codons. F3 and. Then a given amino acid will appear at different ordinate values. 7 Local encoding functions F1 .g. which leads to the representation of each triplet as a sextuple of zeros and ones. We have defined in the set of 64 triplets a structure of a vector space over the binary field Z2 = {0. 1}.6. A graphical representation of the local encoding functions F1 . F3 and F4 70 F1 F2 F3 F4 20 18 19 60 21 20 21 50 17 16 15 17 Representative codon 40 12 11 11 5 10 14 13 30 9 9 8 6 11 20 6 5 7 10 4 2 3 1 4 4 0 0 1 10 20 30 40 50 60 70 Codon (lexicographic order) Fig. Pro appears when F2 and F4 are applied) or even three (e. The general encoding function for amino acids and stop codons in the hypercube of 64 triplets As it is well known. In Table 3. one representative codon but some of them appear in two (e. we derive an algebraic function.230 Bulletin of Mathematical Biology (2007) 69: 215–243 Encoding Function F1. F2 . as a vector of the six-dimensional vector space (Z2 )6 . 3. Ser appears when F1 . that is. also called six-dimensional hypercube. 61 out of the 64 triplets code for the 20 primary amino acids that are the building blocks of proteins. associated to the vectorial structure. A ↔ 01. are applied) of the subsets. the triplets . G ↔ 11. This correspondence involves the addition operation and the vectorial algebraic structure in the set of triplets. It is done by means of the assignment C ↔ 00. F4 . F2 . which assigns to every triplet its associated amino acid or its termination mark. F2. as well as a combinatorial geometry.g.

UCU. there are also two special sets. Thus. The overall shape is still linear and we remark that this function represents the actual SGC. We define the encoding function as that which assigns to every triplet the left-most triplet in every row. the one which is in the left-most position (marked in bold characters). UCA. The first triplet in each row. The latter is mainly due to a change in the first nucleotide of their codons. Table 4 Triplets for which the function F requires an affine transformation Amino acid Arg Gln Glu Leu Lys Met Ser Trp Stop Stop Canonical triplet CGC CAA GAA CUC AAA AUG AGC UGG UAA UAA Special set AGA. UAG UGA Encoding function t2 ◦ F t6 ◦ F t6 ◦ F t1 ◦ F t6 ◦ F t56 ◦ F t1234 ◦ F t56 ◦ F t6 ◦ F t36 ◦ F . a graphical representation built in the same way as Fig. and in the ordinate the value corresponding to the canonical codon for each amino acid or the termination mark (image of the function E) is given. UCG UGG UAA. In fact. In Table 4. Three out of the 19 codons specify the stop signal and the remaining eight codons correspond to the three amino acids encoded by six codons whose canonical representatives are of the type NNC. is taken as the canonical representative of the corresponding amino acid.Bulletin of Mathematical Biology (2007) 69: 215–243 231 which code for each amino acid and the stop signal are listed according to their lexicographic order. Eight out of the 19 codons specify the remaining five amino acids and their canonical codons are of the type NNA or NNG. 7 is shown in Fig. The 45 triplets encode for 15 amino acids. GAG UUA. In the abscissa. we show the encoding functions for every special set. some of which are mapped directly by F but others require a translation. represented by crosses. AAG AUG UCC. AGG CAA. as shown in Table 4. The remaining 19 codons. There are three amino acids encoded by six codons. it coincides for 45 (grey characters) out of the 64 triplets. their canonical codons are of the type NNC. with the desired encoding function denoted by E. values from 0 to 63 were assigned to each codon according to the selected lexicographic order. it turns out that the endomorphism F coincides. In order to summarize the function E. The other 19 triplets (black characters). CAG GAA. are those for which the encoding function E takes the form of an affine transformation. the composition of F with a suitable translation. in most of the cases. Note that for 8 out of the 20 amino acids there are special subsets of the sets of their associated triplets for which F alone is not the encoding function E. that is. UUG AAA. In this encoding function a single canonical codon corresponds to each amino acid in contrast to the above-mentioned four encoding functions. require a particular translation. There are five amino acids and the stop signal whose canonical codons end in A or G and therefore they require specific translations. 8. The 45 codons that specify 15 amino acids which are directly mapped by the linear function F are represented by circles. For the stop signal.

and AAA. the triplet UCC. which is the first in its list. The graphical representation of the general encoding function E of the SGC. For this reason. that is. whose canonical triplets are. 9. we show a graph representation diagram of . Met. The other five amino acids. AUG and UGG. The first three. respectively. GAN and AAN. 8 Plot of the general encoding function E in its two main forms. AAA. the addition of the triplet CCA. In Fig. It codes for Ser. and UGG (Table 3). but this amino acid is already represented by the triplet AGC. GAA. Lys. under the affine transformation t6 ◦ F . The other two triplets. which is not the canonical representative of any amino acid. The graph representation diagram of amino acids and the stop signal Note from Table 3 that the vertexes of the four-dimensional hypercube NNC (Fig. and Trp. GAA. These three triplets are the unitary images of the faces C AN. CAA. CAA. 6) correspond to triplets that code for 15 out of the 20 amino acids. we deleted the vertex with label UCC and its four adjacent edges in the last graph diagram. AUG. the addition of the triplet CCG. As before. which is the image of NNC under the translation t6 . which is the image of NNC under the translation t56 .232 Bulletin of Mathematical Biology (2007) 69: 215–243 General Encoding function E 60 mapped by F mapped by F and a translation 20 18 19 17 21 50 16 15 Canonical codon 40 12 14 13 30 10 9 11 20 6 5 7 8 5 4 10 4 2 3 0 1 0 10 20 30 40 50 60 70 Codon (lexicographic order) Fig. there is only one. Amongst the 16 triplets which are labels of the vertexes. 4. are Gln. which are not represented in the hypercube NNC . under the affine transformation t56 ◦ F . Glu. belong to the hypercube NN A. that is. belong to the hypercube NNG. these two triplets are the unitary images of the faces AU N and UGN.

with an additional vertex. whose canonical representative triplets do not belong to the hypercube NNC . the 20 amino acids. 1995). These three sets are disjoint when they are pairwise compared. codons of the type NYR and the YRN appear. Altogether these three sets comprise 45 triplets which code for 17 out of the .g. five of which correspond to amino acids. Konecny et al. and that the distance between the amino acids Met and Trp is equal to 3.Bulletin of Mathematical Biology (2007) 69: 215–243 233 Fig. nor its adjacent edges. Each of these types corresponds to one set of 16 elements. A graph representation diagram of the 20 amino acids and the stop signal. 9 The phenotypic graph of the 20 amino acids and the stop signal. which represents the stop signal. The canonical RNA code consists of only RNY codons that comprises 16 out of the 64 possible triplets and which codify for eight amino acids. with the addition of six external vertexes. It is a phenotypic graphical image of the hypercube NNC . Discussion In this work. we propose an Extended RNA code as derived from the RNA code as originally proposed by Eigen (1977) and later used by several authors (e. Gln and Tyr. without the vertex UCC. and another vertex.. We observe that the vertex which represents the stop signal is adjacent to the amino acids Glu. By allowing readings starting at the SRF and TRF positions of the RNY code. 5. that represents the class of the three stop codons.

and they also include the three stop codons. Interestingly. The remaining 16 triplets. which we call RNA-less code or complementary code of the Extended RNA code. and they proposed that this order may reflect the evolution of the genetic code from an RNY structure. These 16 triplets constitute two disjoint sets. via the Extended RNA code and the addition of the RNA-less code. and YYY and RRR (RNA-less code) (Fig. some amino . The union of the cubes YYY and RRR produces a four-dimensional vector subspace which is not a hyperface of the six-dimensional hypercube. the steps RNY plus RNR and YNY could also form another extended RNA code. frame-reading mistranslations conferred obviously evolutionary advantages. and YRN (Extended RNA code).2) to each other as affine subspace of the whole sixdimensional hypercube. each of them being a three-dimensional cube. each four-dimensional hypercube is isomorphic and isometric (A6.234 Bulletin of Mathematical Biology (2007) 69: 215–243 20 amino acids. we can hypothesize that the point in which genetically encoded protein translation started to evolve corresponds most likely to a breakthrough organism obeying an Extended RNA code after the RNA World and prior to the cenancestor. this subspace is isomorphic as affine subspace to the three hypercubes of the Extended RNA code but it is not isometric to any of them. Our present results do not offer any clue about a chronological order in which the different encoding subsets could have led to the current SGC. 32 new triplets (which codify for nine new amino acids) and three stop triplets (which specify a stop signal). we can decompose the six-dimensional hypercube as consisting of the patterns RNY (primaeval RNA code). can not be derived by frame-shift readings but rather by other types of mutations such as insertions. emerge.. Conversely. Alternatively. It works on what already exists. Given the RNA code. 1985). since in fact. the union of the three hypercubes of the Extended RNA code and the vector subspace of the RNA-less code. In the RNY code every amino acid is coded by two neighbor triplets located in an edge whereas in the NYR and YRN codes there are departures from this regularity: some amino acids are now encoded by four triplets and others are encoded by only one. The RNY code can be graphically represented as a four-dimensional hypercube that results from the union of the disjoint sets RYY and RRY each of them being a three-dimensional cube. respectively. It innovates with what it has at hand and this process has been recognized as the evolutionary tinker (Jacob. plus NYR. These results suggest a plausible evolutionary path from which the primaeval RNA code could have originated. each of them being a three-dimensional cube. As a consequence. and they are pairwise disjoint. and the union of the disjoint sets YRY and YRR. However. Notably. The NYR and YRN sets are also represented by a fourdimensional hypercube that result from the union of the disjoint sets YYR and RYR. Natural selection does not generate novelties from scratch. it has been found that the order of triplet frequencies RNY > RNR > YNY > YNR is a general attribute of coding sequences (Eigen et al. This is what we call the Extended RNA code. the six-dimensional hypercube. makes up the whole six-dimensional hypercube of 64 triplets. 1977). YYY and RRR. deletions and substitutions. 5). by allowing reading slippages in the other two reading frames. Thus. providing a comma-free readout via wobbleintermediates to the present form. either transforming a system to give it new functions or combining several systems to produce a more elaborate one.

´ 1999. respectively.1. 1990). and with that of vector subspace of a vector space. Given the uneven degeneracy of the genetic code it is appealing that the general encoding function is almost linear.2. to every subset of the form v + W. to any of the three four-dimensional hypercubes of the Extended RNA code.1. a single mRNA can be translated in three different reading frames which encode three trans-activators that are required for late transcription (Keck et al.. A.1. Konecny et al. and it contains the element v . It is also worth to mention that in present day DNA virus such as vaccinia virus.Bulletin of Mathematical Biology (2007) 69: 215–243 235 acids of the RNY are also coded by triplets that appear in the SRF. et al. Remarks.1. In the vaccinia virus. for a linear variety. The associated vector subspace W. as well as with the concept of a linear transformation or linear endomorphism of a vector space. this is the first time in which the SGC is expressed as a mathematical function which maps each triplet onto its corresponding amino acid or stop signal.. Appendix A. but the vector v may be any of the elements of the set v + W. a single messenger RNA (mRNA) is able to encode three different proteins because messages contain three distinct putative translation initiation sites. In the context of the frozen concept. (1995) first noted that by allowing reading slippages there were two hidden messages in the RNY code which are AUG and CAU which are found in the SRF and TRF. Mathematical and biological background We assume that the reader is familiar with the concept of a vector space over a scalar field K. is unique. where W is a vector subspace and v is a fixed vector. In the search of producing synthetic life in the laboratory (Hutchinson III et al. and there are new triplets which altogether specify nine new amino acids. in part. We recall that every vector space is an abelian group for the addition operation. 1990). we can say that considering the symmetries of both the Extended RNA code and the RNA-less code. For a vector space V we call affine subspace of V . the primaeval RNY code was already frozen and that it evolved like a replicating and growing icicle. also called a K-vector space. or linear variety contained in V . In other words. The phenotypic graph of amino acids is also a novel finding whose image resembles. The set v + W is also called a coset or adjoint class of the subgroup W. To our knowledge. The dimension of a linear variety is defined as the dimension of its associated vector subspace. When two vectors u and v define the same linear variety. genes are transcribed in a frame-shift fashion (Keck. for a fixed subspace W. it means that their .. Definition A. The subspace W is called the associated vector subspace of the linear variety v + W. 2005) our encoding functions may be used as a guide to understand the difference between a tinkered-together genome and an engineered one. Szathmary. Concept of affine space and its dimension Definition A.

In particular if W = U . the set or family of all the affine subspaces or linear varieties contained in V .2. . The concept of parallelism Definition A. xn ). . defined as the set of all the ordered pairs (u. . . provided of a coordinate system. en = (0. . Consequently.. n}. . .. or U is a subspace of W. From standard courses of linear algebra.2. We say that 2 affine subspaces v + W and u + U are parallel if W is a subspace of U .3. The affine subspaces of dimension 1 are called the lines of the geometry and those of dimension 2 are called the planes of the geometry. . the vectors are represented as n-tuples (x1 . The only affine subspace of dimension n is the whole space V . or as some vector which belongs to W. are called the coordinates of the vector. every n-dimensional K-vector space is isomorphic to the space Kn . also called points of the associated geometry. . 0. if n is its dimension. The concept of n-dimensional hypercube Let us consider a vector space of the form V = Kn . etc. where K is a field. Definition A. the affine subspaces that contain the null vector. 0. is the ordered set of vectors e1 = (1. A. . and plane in a geometry. The elements xi . they form a base of the vector spaces. . 0).2. The isomorphism may be defined by the matching of any of the bases of V with the canonical bases of Kn . v ) ∈ V × V . The linear varieties or affine subspaces. In this case.2. The other affine subspaces. Definition A. According to this definition we can talk of parallel lines. If we take v as the null vector 0. respectively. for a fixed linear subspace W.2. 1. the affine subspaces of dimension 0. K being a field. lines and planes in a geometry are. . for i ∈ {1. in an n-dimensional vector space V . As it is well known. they are parallel. a line parallel to a plane. parallel planes. . are the equivalent classes of the equivalence relation RW . The canonical base of Kn . all the vector subspaces are also affine subspaces. 0). the k-dimensional for 2 < k < n. 0.1. x2 . parallel cubes. if the affine subspaces have the same associated subspaces. the identity v + W = W is obtained. it is known that every affine subspace v + W.236 Bulletin of Mathematical Biology (2007) 69: 215–243 difference vector u − v belongs to W. such that u − v ∈ W. are generically called hyperplanes of the geometry.3. is the solution set of a linear system . A. Hence. line. . . 2. . Thus. . all the unitary sets are affine subspaces of dimension zero. where the xi are elements of K. 1. . . The points. e2 = (0. and 2. . having the same dimension. 0. that is. that is. 1) . We call the associated geometry of a K-vector space V . . Then we say that the space has been coordinatized. that is. . It is easy to show that these vectors are linearly independent and that they generate the space. The associated geometry to a vector space. Concepts of point. in fact.

The concept of coordinated affine subspaces Definition A.Bulletin of Mathematical Biology (2007) 69: 215–243 237 of n equations with n unknowns.1. being the linear subspace.2.4. the norm is simply the number of ones that appear in the n-tuple.e. In Fig. a coordinated affine subspace is the solution set of a system whose associated matrix is diagonal. . represent the points which are the vertexes of a cube or regular hexahedron.. The vectorial coordinated lines are usually called coordinated axes and the vectorial coordinated planes simply coordinated planes.1. 1973). . of two elements. which is always a nonnegative integer. the solution set of the associated homogeneous system. the vector space Kn is called n-dimensional hypercube. The name hypercube comes from analogy with the three-dimensional case. where the eight triplets of zeros and ones.5. A coordinated affine subspace of the hypercube is called a hyperface of the hypercube (Z2 )6 . and all the edges have length 1. coincide with the extremes of the vectors e1 .. The two-dimensional hyperfaces are simply the faces of the hypercube. In the particular case where K is the binary field Z2 = {0. its defect remainder in the entire division by p. We will call norm or length of a vector v = (x1 . where (Z p ) denotes the Galois Field of p elements. being p a prime number. 1}.3. 0. . the six-dimensional hypercube of the 64 codons is illustrated. which are remainders of the entire division by p. it is also called the weight of the vector. Definition A.e. In this case. one of whose vertexes is the null vector 0 = (0.4. the field of non-negative integers. The ndimensional hypercube (Z2 )n is a regular polytope (Coxeter.5. i. The dimension of W is equal to n − r . Definition A. where r denotes the rank of the associated matrix. The concept of norm or length of a vector in the spaces (Z p )n Let us consider a vector space of the form (Z p )n . x2 . 0). In the case of the binary field (Z2 ). The zero-dimensional hyperfaces are the vertexes of the hypercube whereas the one-dimensional hyperfaces are the edges of it. In fact. . the system with the same left parts and the right members equal to zero. It is also called an orthotope because the angle of two adjacent edges is a right angle. is the so-called reduction module p. i. A. xn ). the 2 2 2 ordinary sum |v | = x1 + x2 + · · · + xn . A. The assignment to every integer.4. Definition A. In a vector space Kn we will call coordinated affine subspace to every affine subspace whose associated subspace is generated by one or some vectors of the canonical base.1. e2 and e3 of the canonical bases. . and the vertexes which are adjacent to it. 5 of the main text.

have the property that the inverse of any of them is equal to its transposed. this distance is the so-called Hamming distance. A. A transformation f of the vector space V = (Z2 )n in itself. .e. Note that the norm or length of a vector coincides with the scalar product of the vector with itself. and geometrically. Definition A. we can define the concept of scalar or inner product in the following way: Definition A. . visualizing the space as a graph. In general. v ) of arbitrary vectors u and v . In vector spaces of the type (Z2 )n . v . It is easy to show that a linear isometry preserves the length or norm of every element of the space.1. We define the distance between the vectors v and w as the norm of the difference vector v − w . . called permutation matrices. every linear isometry performs a permutation on the set {e1 . . .5. We call scalar or inner product of the vectors v = (x1 .2. For this reason. y2 . .2. the so-called hypercubes. the Hamming distance is the number of places in which both vectors differ. i.7. widely used in coding theory and in criptology. .6. . xn ) and w = ( y1 . The concept of permutation transform or multiple rotation in the vector space V = (Z p )n As the only vectors of length 1 in the vector space are the canonical vectors e1 . . An isometric linear transformation will be called a linear isometry of the vector space. In the case of the hypercube (Z2 )n . f (v )) = d(u. the affine coordinated subspaces are called hyperfaces of the hypercube. . Actually.6. For every pair (u. . It can be proved that the inner product and the length or norm are related in the following way. the linear isometries will also be called permutation transforms. e2 . is called isometric if it preserves the distance between every two elements. en }. The matrix of a linear isometry with respect to the canonical base is a matrix obtained from the identity matrix by a permutation of its columns. . Transformations that preserve the distance Definition A. e2 . v ) for every pair (u. if d( f (u). A.6. .. it means the minimal number of edges between the two vertexes represented by the n-tuples. the affine coordinated lines are usually called edges..238 Bulletin of Mathematical Biology (2007) 69: 215–243 In the more general case of a vector space (Z p )n . yn ) the integer < v. v ) ∈ V × V . being the subtraction the inverse operation of the addition in the vector space (Z p )n . the following equality holds: |u + v | = |u| + |v | + 2 u. and the affine coordinated planes are called faces of the hypercube. x2 . en . . The concept of distance in the spaces (Z p )n . . w >= x1 y1 + x2 y2 + · · · + xn yn . This kind of matrices.

A. a multiple rotation in (Z p )n can also be interpreted as a multiple rotation in Rn .7.1. We call affine transformation to every composed function of the form tv ◦ F . being R the field of real numbers. The concept of translation and affine transformations in a vector space Definition A. . a linear isometry preserves the orthogonality of any pair of orthogonal vectors of the vector space. T is. It means that the set T of all the translations of the vector space V is a group. and so on. being the isomorphism. For translations tv and tw . According to this definition the vectors of the canonical bases. associated to the vectors v and w . As the set (Z p )n can be viewed as a subset of the Rvector space Rn . ti jk as the translation associated to the vector ei + e j + ek.9. additionally. i. every linear isometry can be interpreted as a composition of local rotations. The linear isometries of the vector space (Z p )n will also be called permutation transforms or multiple rotations.8. Definition A. We say that two vectors v and w of the space (Z p )n are orthogonal or perpendicular. and they are. Special notation. its linear component F is also bijective. The concept of orthogonality in a vector space (Z p )n Definition A. from V onto T . Definition A. as ti j .e. in fact. Hence. In the hypercube (Z2 )n . Given a vector v of a vector space V . It is clear that an affine transformation tv ◦ F is bijective. to the bijective function tv : u → u + v . associated to the sum v + w of both vectors. k ∈ {1.1. if their scalar product is equal to zero.1. one to each other. This kind of base is usually called orthonormal base. If the affine transformation is bijective.2. an abelian group which is a subgroup of the symmetric group S(V ) of all the bijections of the set V with itself. where i . A. the translation associated to the vector ei + e j .9. and only in this case. we will denote as ti the translation associated to the vector ei of the canonical base.8. It can be proved that a linear isometry preserves not only the distance between vectors. n}.. of length 1. but also the scalar product of any two vectors of the vector space. . in any vector space of the form (Z p )n are orthogonal. one to one. . given by the function: v → tv . whose dimension is less than or equal to that of v + W. .9. 2. and only if. a multiple rotation in the vector space. . from V to V .Bulletin of Mathematical Biology (2007) 69: 215–243 239 As every permutation is a composition of pairwise disjoint cycles. the dimension is preserved. that is. where F is a linear transformation or linear endomorphism of the vector space and tv is the associated translation of a fixed vector v . isomorphic to the additive group of V . It is easy to prove that every affine transformation carries a linear variety v + W onto another linear variety. if. j . the composition tv ◦ tw is the translation tv+w . we call a translation associated to the vector v .

If the translation is to a distance equal to 1. that is. 0. and G= 11. since it consists in the change of only one base. if the triplet CUA changes to the triplet CUG.240 Bulletin of Mathematical Biology (2007) 69: 215–243 If v + W is a k-dimensional hyperface and u is any vector of the hypercube (Z2 )n . by substitution of the base A by the base G. The substitution in a triplet of one or more of its three nucleotide bases within the set {C. associated to some vector v . that is. is an affine subspace. u. It can be proved that the union of the k-dimensional hyperfaces v + W and ei + v + W.10. it is a transversion. when u is orthogonal to W and has length h greater than 1. Transitions and transversions. that every triplet XYZ is represented as a sextuple (x . 0. is a (k + 1)-dimensional hyperface. Definition A. The nucleotides A and G are called purines. 1. to a distance h greater than 1. associated to it. 0. z. U= 10. Mutations Definition A. A= 01. that is. y. Definition A. A right hyperprism is obtained from a hyperface by the adjunction of a translation of it. 0. when ei does not belong to W. 0. . In this particular case the mutation is simple. 0. and it is called a transversion when the change is from one class to another. if the triplet CUA is changed for the triplet CUG. for instance. it means that the vector (0. The union of the hyperfaces v + W and u + v + W. 0. On the the other hand. 0). the nucleotide A is replaced by the nucleotide G. t .3. A. then the purine A is changed to the pyrimidine U. with h greater than 1. but it is not a hyperface of the hypercube.9. U. as an element of the hypercube (Z2 )6 . 0. We say that the mutation is simple if the substitution involves only one base of the triplet. 1) and this transformation may be represented by the addition of the vector e5 = (0. A right hyperprism of height h.2. For instance. G} is called a mutation. that is. 1. every mutation may be represented by means of a translation tv . 1. and the nucleotides C and U are called pyrimidines. if CUA is changed to CUU. purine to pyrimidine or pyrimidine to purine. by the translation t5 . According to our representation of triplets as elements of the vector space (Z2 )6 . then the set u + v + W is also a k-dimensional hyperface.10. Then. a (k + 1)-dimensional hyperface is obtained. Then. A codon mutation is called a transition when it changes a base to a base of the same class.10. 1. v ) of zeros and ones. This terminology is due to their chemical composition. since A and G are both purines.1. 1) is changed to the vector (0. purine to purine or pyrimidine to pyrimidine. In our work. It follows. we have assigned to every nucleotide a numerical pair in the following way: C= 00. is called a right hyperprism of height h. it is a transition. Some biological concepts and their mathematical representation A.

and Z are the complements of X. t15 . defined by the vector M.12. are obtained by additions of the triplets ACC. one to each other. Y . it means the addition of the triplet GGG to XYZ. t3 . which have odd subindexes. the maximum of all the possible distances in the six-dimensional hypercube. y. 1. CAC and CCA (see Table 2). t4 . The mentioned function is the translation t M . where the components X . Notice that the Hamming distance between a codon and its complementary is always equal to 6. u. that is. which assigns to every triplet XYZ the triplet X Y Z . v ). t13 . respectively. as additions of the triplets UCC. In the vector space of the 64 triplets we can define a bijective function. that is. the base A is paired with T in the double helix structure of DNA. and the identity function may be considered as a transition. and t6 of even subindexes. t3 and t5 . which is the sum e1 + e2 + e3 + e4 + e5 + e6 . From the algebraic point of view. the set of all the transitions is a group of order 8. the vector M = (1. t35 and t135 being t0 the identity function. t3 . the associated to the vector M. and it consists in the substitution of every zero by ones and every one by zeros. They also form a pair in the double helix structure. CUC and CCU. the Hamming distance between a base and its complementary is always equal to 2. those associated to the canonical vectors e1 . that is. Analogously. the addition to the vector (x . whereas the basic transversions represented by the translations t2 . z. of the six vectors of the canonical base. the subgroup of pure transitions is a normal subgroup of T . The set of mutations which are transitions is closed under the composition. t5 . A. In the Boolean structure of the hypercube (Z2 )6 this transformation is the so-called Boolean negation.11. according to the addition operation illustrated in Table 2. they have been represented by complementary numerical pairs 10 and 01. while the basic transversions are represented by the translations t2 . For this reason. t . subgroup of the group T of 64 possible mutations. The sum of a triplet with its complementary gives.Bulletin of Mathematical Biology (2007) 69: 215–243 241 A. . the triplet GGG. As T is an abelian group. and t6 . 1. and t5 . represented by the translations t1 . e3 and e5 . This triplet is usually called the complementary codon or the anticodon of XYZ. According to our numerical representation. 1. 1). The eight transitions are represented by the translations: t0 . and Z. for they differ in two of their components. respectively. as result. and are represented by the numerical pairs 00 and 11. are obtained. Y. the pyrimidine C and the purine G. the purine A and the pyrimidine U are considered to be complement of each other. The basic transitions. The translation t M performs a transversion in each of the three components of the triplet. t1 . t4 . Complementary bases For biological reasons. are considered complementary bases. Hence. associated to the triplet. Algebraic representations of transitions and transversions Notice that the basic transitions in the hypercube (Z2 )6 are represented by the translations t1 . All the other mutations are transversions or compositions of transversions with transitions. 1. When U is changed by T (thymine). that is.

Regular Polytopes. Silva. Hutchinson III. and time. The hypercycle. Eigen. Govezensky. Science 196. Lazcano. Biol. Theor... A. M. We also thank Rafael Camacho for helpful suggestions. . Proc. D. Grau. Natl. S. A. Statistical Garc´ ıa. Sabater-Munoz. Nature 319. The search for missing links between self-replicating nucleic acids and the RNA world. 341–369. S. Morgado. S. E. 2004.. Miller. C. A. J. Ricci. Peterson.. 52. 1999.. 1405–1421. M. R. 66.. 1978.. 1973.. 801–809. Koonin. Biosph.R. Orgel. S. Schoniger.R. 1995. Hurst.. 25. E... Silva. 29–46.. Mol. Biosystems 39. Mushegian.. 618. Jose. White. Sci. Proc. F. J.I.. 1161–1166. ˜ ¨ Jimenez-Monta no. The hypercycle. Math. Keck.V. E.V. Sci. 1957. Griffith. 1997.: Toward the minimal genome needed for symbiotic life. 1986. The RNA World. Natl. Acad.G. Cech. 10268–10273.. Complementary coding to the primeval commaless code. 793–798. 1999. Cline.D. Jacob. Venter. T. Dover Publication Inc. 173. Gesteland. Eigen.. Atkins. F. Evolution and tinkering.. L. Cellular evolution during the early Archean: What happened between the progenote and the cenancestor? Microbiologia SEM 11. Gill. Cuba.. 45. J.. 1990.. Naturwissenschaften 64. 263–270. Bobadilla. The origin and early evolution of life: Prebiotic chemistry. 1977. and by the Mul´ y la Computacion. Santa Clara. Cold Spring Harbor Laboratory Press.. Clarke. 367–379. L. Flores..H. Mol. Kenneth. Baldick. J. 1968.. 2002.J.J. Petoukhov.F.C...T. S. analysis of the distribution of amino acids in Borrelia burgdorferi genome under different genetic codes. J.. W. Hofacker. Biol. P. A principle of natural self-organization. MATCH Commun. Naturwissenschaften 64. E....H. Smith. Leguina..D. USA 43.. A. 2004. C. Coxeter. Acad. T.E. M. F. R... P. The genetic code Boolean lattice.. B. A. 1995. Natl. Acad. Moya.. New York.. Lazcano.. Ellington. Alvarez. 38. Life Evol. A. Hamming distance and stochastic matrices. References Becerra. J.A. Part C: The realisitic hypercycle.. 1996.. Physica A 342. 1977..M. de la Mora-Basanez. Proc. The genetic code is one in a million. Biol. S.R.. 515–530. Evol. The RNA World. Freeland. Moss. Poschel.. ´ M. R. M. Sci. 1996.O.H. ´ Sanchez. S. Part A: Emergence of the hypercycle..E.R.C.S. UNAM. A principle of natural self-organization. 417–421. M.F. ´ tiproject: Tecnolog´ ıas para la Universidad de la Informacion ´ UNAM. A minimal gene set for cellular life derived by comparison of complete bacterial genomes. B..S. P. 4454–4458.. Latorre. J.. Mexico.V.. He. C. Extreme genome reduction in Buchnera spp. A. Codes without commas. H. 2437–2441. was supported by Universidad Central ”Marta Abreu” de ˜o ´ for preparing the Las Villas. Orig. 115–118. M.242 Bulletin of Mathematical Biology (2007) 69: 215–243 Acknowledgments M. J.V.. USA 99. H.J. B. Role of DNA replication in Vaccinia virus gene expression: A naked template is required for transcription of three late trans-activator genes. 2165–2169..R.A. Schuster. A. R. Lazcano...J.A. J. USA 93. The hypercube structure of the genetic code explains conservative and non-conservative amino acid substitutions in vivo and in vitro. ´ ˜ M. Pattern analysis of 5S rRNA.. C. Natl.. ¨ Konecny. Math.. The origin of the genetic code.. Science 286.. Crick. 541–565. Acad.. Eigen. We thank Giselle Morgado Avin figures of the manuscript. Proc. Comput. 117–125. 2004a. L. ˜ Gil... T. the pre-RNA World. 1995.. Schuster. Crick. Global transposon mutagenesis and a minimal mycoplasma genome. Islas. USA 82. J. Winkler-Oswatitsch.L. Chem. R.C.N. J. was financially supported by PAPIIT-IN216106. Genetic code. 238–248. 1996.. R.. Cell 85. Sci. 47. Cell 61. 3rd edition. Gilbert. Evol. C. 1–13. 1998. 1985. Mol. Bull.J. J. S. O.. 288–293. A... Polyphyletic gene losses can bias backtrack characterizations of the cenancestor.. Lindemann. Fraser..

R. 190–197.. 67. Biol. ´ Sanchez. E. . R. White. 1. 2004b....com/Images/genetic/Max/Sym300pi. In search of the simplest cell. Maximum symmetry in the genetic code: The Rafiki map.. E. Corfus.pdf. The meaning of boolean deductions. 1967. New York. 433.. Ch. ISNN 1109–9518. 469–470. Genetic code Boolean algebras. vol. Morgado. Woese. Unpublished manuscript. The Genetic Code. issue 2. 1–14. 7.. 2005. E..codefun. Bull. 2005. Science. http://www. Greece. Undated.Bulletin of Mathematical Biology (2007) 69: 215–243 243 ´ Sanchez. Grau. R. M. R. pp. C. Math. Grau. Harper and Row. Morgado. A genetic code boolean structure I. W-Seas Transactions of the International Conference on Biology and Biomedicine. ´ Szathmary..

Sign up to vote on this title
UsefulNot useful