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).

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

is defined in Table 1. 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 . according to the terminology of Coxeter (1973) (A3). This group is isomorphic (A3) to the group of all the symmetries of a plane rectangle (not including the square). (G. which is the operation table of a known abelian group of order 4. U. A). This is so because the picture is a projection of a six-dimensional figure over a plane. A).pyrimidine. Y . This sum. G. Results 2. C). The 64 sextuples of zeros and ones generate the so-called six-dimensional hypercube. In the hypercube. G. 2. and (G. U. The same happens with some of the angles of a three-dimensional cube. In the vector space structure of the hypercube the addition is the so-called sum module 2. The RNA code and its three reading frames: The extended code A primaeval RNA World was proposed (Gilbert. 1}(A3. but in our pictures many of them look like acute or obtuse. the so-called Klein Four Group. Since the primitive translational apparatus may have been imperfect. A. When Table 1 is translated to the nucleotide bases it gives Table 2.218 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. as in every orthotope. but it is possible to show that the remaining ones lead to the same results. From these orderings. all the angles between adjacent edges are supposed to be right angles. for the elements of the field Z2 . C.purine. shifts in the reading frame probably occurred.any nucleotide). and codons with an NYR and YRN patterns emerged. These three patterns altogether are here defined as the codons of the Extended RNA code. The vector space (Z2 )6 is an orthotope.1). C. 1986) which comprises the codons with an RNY pattern (R . (U. bit by bit.1. A. N . which is a vector space over the binary field Z2 = {0. we have selected the first. also called XOR operation. C). widely used in symbolic logic. when they are portrayed in a plane.

that is. that is. Finally. From the eight subsets. that . j ). 1b) in which every amino acid corresponds to an edge of the associated cube. we introduce the following notational convention.2) in the first nucleotide. and the three known stop codons. In this context.. but also the same first nucleotide. RRY is obtained from YYY by the addition of the triplet AAC. including AUG. we denote as ti the associated translation: v → v + ei . This is so since the Hamming distance (He et al. The set of codons of the form RNY is the union of the cubes RYY (Fig. k). these subclasses correspond to coordinated planes or faces of the cube.1) between triplets on the same edge is 1. 1a) and RRY (Fig. For every triplet (i . j . which represents transversions in the first and the second nucleotides. and we defined it as the RNA-less code or complementary code of the Extended RNA code. of the canonical base. In most cases. not only the same central nucleotide. respectively. 2004. In order to describe algebraically and geometrically all sets. A. the start codon. The first set corresponds to the so-called primaeval RNA code (RNY pattern). The other seven subsets are three-dimensional affine subspaces obtained from YYY by translations (A9. each of the subclasses is partitioned into two pairs. The eight elements of a subset correspond to a three-dimensional coordinated affine subspace (A4. 1968. we may consider two subclasses. for the set RNY we obtain the two subsets RYY and RRY. For every pair (i . an ordinary cube. The remaining 16 codons can only be obtained by mutations other than by frame-shift readings.1). 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. 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. G}. Eigen and Schuster. 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). and Z belong to the set {C. which contains the null vector CCC . 1978). For each of the eight subsets. the addition of the vector e2 which involves a transversion (A10. 2. where X. and that is the subset YYY. every pair consisting of triplets that have.1). we denote as ti jk the composed function ti ◦ t j ◦ tk and so on (A9. A6.Bulletin of Mathematical Biology (2007) 69: 215–243 219 acids. these pairs of codons specify the same amino acid. Now we will consider the combinatorial geometry. and the fourth set corresponds to the RNA-less code (YYY and RRR patterns). The fourth set is the union of two subsets: YYY and RRR. contained in the six-dimensional hypercube.1) or vectorial cube. Each of the first three sets can be partitioned into two subsets by replacing the N by Y or R. U. Each cube is obtained from the other by means of the translation t4 . can be partitioned into four sets of 16 triplets each. Y. that is. associated to the canonical vector e4 . For every vector ei . only one is a vector subspace (A1. For example. RYY is obtained from YYY by the addition of the triplet ACC. the vector e2 + e4 . They constitute coordinated lines or edges of the hypercube. First reading frame The 16 triplets of the RNY pattern code for eight amino acids and are considered as the primaeval RNY code (Crick. associated to the structure of the vector space. we denote as ti j the composed function ti ◦ t j .1). each of them formed by triplets having the same central nucleotide.2. the 64 triplets XYZ of the SGC.

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

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

YRR is obtained from YYY by the addition of the triplet CAA. We will consider the union of the sets of codons of the form RNY. which is a coordinated affine subspace. where 45 out of them specify 17 of the primary amino acids. associated to the canonical vector e2 . by the addition of the codon ACC to every triplet. It means that the set of codons of the form YRN constitutes also a four-dimensional hypercube. or double rotation. those associated to the first. In summary. NYR and YRN. 3b). e4 → e2 . which is a coordinated affine subspace. second and third reading frames. The Hamming distance between a vertex and its image. the addition of the vector e4 + e6 which involves transversions in the second and third nucleotides. ignoring the first (R) and the second (N) nucleotides. e6 → e4 . e4 → e2 . or four-dimensional hyperface of the six-dimensional hypercube of the 64 triplets (Fig. or a four-dimensional hyperface of the six-dimensional hypercube of the 64 triplets (Fig. the set of 48 codons of the form RNY. that is. the triplets of the form YRN give rise to six new amino acids. Then. 3c). due to a slippage. and it can also be interpreted as a double rotation in the six-dimensional Euclidean vector space R6 (A7). that is. under the translation t2 . e3 → e1 . The matrix of this linear transformation is orthogonal with determinant 1. which complete a set of 17 out of the 20 primary amino acids. then triplets of the form YRN will be translated. 2. Thirteen out of the 16 triplets of the form YRN code for six new amino acids. which is the same basis permutation. The hypercube YRN is the image of the NYR under the non-singular linear transformation e1 → e5 . is equal to 1. and their correspondence with 17 amino acids and stop codons. It means that the set of codons of the form NYR constitutes also a four-dimensional hypercube. e3 → e1 . YRY is obtained from YYY by addition of the codon CAC. that is. e6 → e4 . that is. Each cube is obtained from the other by the translation t6 . and the other three codons correspond to a termination signal. e2 → e6 . of the vector e4 that represents a transversion in the second nucleotide. The third reading frame If in a sequence of triplets of the form RNY. is equal to 1. 3a) and YRR (Fig. which is defined by an even permutation of the canonical base. The set of codons of the form YRN is the union of the cubes YRY (Fig. without repetitions of any of those found in the FRF or the SRF. under the translation t6 . e5 → e3 . 2c). that is. 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. The hypercube NYR is the image of the RNY under the non-singular linear transformation e1 → e5 . The other three codons are the so-called stop codons. . NYR and YRN. associated to the canonical vector e6 . The Hamming distance between a vertex and its image. e5 → e3 .4. e2 → e6 . 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 . This function maps every triplet XYZ onto the triplet YZX. The set of triplets of the form YRN is a disjoint set with the sets RNY and NYR. which converts RNY into NYR. as an extension of the primaeval RNY code. comprises the Extended RNA code. 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. the reading starts at the third nucleotide Y.

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

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

Encoding functions So far. the six´ ˜ et al. but it is not as in the other cases. 3. black • YYY and RRR (RNA-less code). It means that the set of codons of the RNA-less code do not lead to a four-dimensional hypercube. under the translation t246 . 5 A graph representation diagram of the Boole lattice of the 64 triplets of the SGC.. purple • YRN (RNA code in the TRF). 5). it has been customarily to represent in various graphical ways (e. The sixdimensional hypercube can be envisaged as composed by: yellow • RNY (primaeval RNA 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.3). a four-dimensional hyperface but rather it is a right hyperprism of height 3 (Fig.g. The set is a fourdimensional subspace of the six-dimensional hypercube. is equal to 3.Bulletin of Mathematical Biology (2007) 69: 215–243 225 image. 1996). none of these Fig. Interestingly. the icosahedron or dodecdimensional hypercube (Jimenez-Monta no ahedron (White undated). orange • NYR (RNA code in the SRF). 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. Actually. . 4c and A9.

denoted by Im( F ). UGA UGG AAA. GAU GGC. We select the representation of every amino acid by only one triplet. UUG AUG CAC. 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. y. CUU. or endomorphism F (A9. GAG UUC. CUA. ACG AUC. The image of F . GGA. U. the triplets of the form CCZ. ACA.1. of the vector space (Z2 )6 . The ordering of the set of triplets is the linear order determined by the selected order of the bases (C. it changes the third nucleotide by the nucleotide C. that is. The kernel of F is the two-dimensional subspace of vectors of the form (0. the RNA-less code. in fact. the first.GGU. which carries over every triplet XYZ to the triplet XYC. 0. the correspondence of the ordered set of triplets that specify every amino acid is illustrated. UGU CAA. GCU.CCA. An auxiliary linear function for the different encoding functions In Table 3. F : (x .UCG GAC. y. GUG AAC. that is. UAG. AAG GAA.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. those that end with the nucleotide C. CAU CGC. GGG CCC. u. z. which are called projections and are characterized by the property of idempotency. CCG CUC. A. 3. v ). t . is the set of triplets of the form XYC. Consequently. GUA. Hence.UUA. AAU AGC. Biologically. F ◦ F = F . CGG. the representative of Ala is the triplet GCC. CUG. AUU GCC. UCC. we derive encoding functions for the Extended RNA code. GCG GUC. ACU. It is a four-dimensional subspace. 0). CCU. t . 0. that is. AGC UAC. AUA. and the SGC. For the three stop codons we select as representative the triplet UAA. UCU. 0. that is. CGU. CGA. AGU. Herein. The endomorphism F belongs to a class of endomorphisms. GUU. G) so that the triplet at the left-most position is the one with the minimum value.2). For example. UCA. GCA. AGA. UAU UGC. a vectorial four-dimensional hypercube or a hyperface of the six-dimensional hypercube. u. CAG UAA. UUU representations is the genetic code. 0. It is the solution subspace of the homogeneous linear . we consider the linear transformation. the kernel of F is a face of the hypercube. z. v ) → (x .

in fact. system u = 0. This set is a four-dimensional coordinated vector subspace. image of the function F1 in the set RNY. the third one is changed to C if it is U. G}. e3 . u. Note that the cube RNC . The encoding function for the primaeval RNY code In the four-dimensional hypercube of codons of the form RNY. or else while preserving the first and second nucleotides. e4 . which projects the whole hypercube onto the cube RNC . which projects it onto its subspace of triplets of the form NNC. a four-dimensional hyperface of the whole six-dimensional hypercube generated by the canonical vectors e1 . is the intersection of the four-dimensional hypercubes RNY . The triplet that is selected as the canonical representative of each amino acid is the one which belongs to the cube RNC . Note that the blue three-dimensional cube is image of F1 . e2 . representing the FRF of the primaeval RNA code. is the minor. to the hypercube RNY. is selected. We denote this hypercube as NNC (Fig. In other words. y. t . v = 0.2. The function F1 is the restriction. the function F1 assigns to each of the 16 triplets of the set RNY the same triplet if it ends with C. v which are the components of a generic vector in (Z2 )6 . Hypercube NNC image of the function F . U. image of the function F . its vertex that ends with the nucleotide C. according to the linear order of the set of triplets as derived from the selected ordering in the set {C.Bulletin of Mathematical Biology (2007) 69: 215–243 227 Fig. A. 6). z. This representative triplet in the cube RNC . we define a function F1 . of the linear transformation F : XYZ → XYC . with unknowns x . if compared with the other representative. 6 The hypercube NNC . of the whole six-dimensional hypercube. 3. This means that for every edge associated to the same amino acid.

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

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

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

a graphical representation built in the same way as Fig. UUG AAA. there are also two special sets. There are five amino acids and the stop signal whose canonical codons end in A or G and therefore they require specific translations. UCA. The first triplet in each row. The 45 codons that specify 15 amino acids which are directly mapped by the linear function F are represented by circles. Thus. we show the encoding functions for every special set. GAG UUA. In fact. In the abscissa. 7 is shown in Fig. The other 19 triplets (black characters). some of which are mapped directly by F but others require a translation. require a particular translation. UCG UGG UAA. In this encoding function a single canonical codon corresponds to each amino acid in contrast to the above-mentioned four encoding functions. 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. UCU. UAG UGA Encoding function t2 ◦ F t6 ◦ F t6 ◦ F t1 ◦ F t6 ◦ F t56 ◦ F t1234 ◦ F t56 ◦ F t6 ◦ F t36 ◦ F .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. AGG CAA. The overall shape is still linear and we remark that this function represents the actual SGC. In order to summarize the function E. The latter is mainly due to a change in the first nucleotide of their codons. values from 0 to 63 were assigned to each codon according to the selected lexicographic order. it turns out that the endomorphism F coincides. their canonical codons are of the type NNC. CAG GAA. 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. For the stop signal. In Table 4. it coincides for 45 (grey characters) out of the 64 triplets. 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. AAG AUG UCC. is taken as the canonical representative of the corresponding amino acid. There are three amino acids encoded by six codons. The 45 triplets encode for 15 amino acids. the one which is in the left-most position (marked in bold characters). We define the encoding function as that which assigns to every triplet the left-most triplet in every row. in most of the cases. represented by crosses. with the desired encoding function denoted by E. that is. The remaining 19 codons. are those for which the encoding function E takes the form of an affine transformation. 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. 8. as shown in Table 4. the composition of F with a suitable translation. Eight out of the 19 codons specify the remaining five amino acids and their canonical codons are of the type NNA or NNG.

the addition of the triplet CCA. The first three. CAA. 6) correspond to triplets that code for 15 out of the 20 amino acids. 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. Lys. which is the image of NNC under the translation t6 . Met. the triplet UCC. which is the first in its list. AUG and UGG. and UGG (Table 3). we show a graph representation diagram of .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. The graphical representation of the general encoding function E of the SGC. belong to the hypercube NNG. These three triplets are the unitary images of the faces C AN. The other two triplets. AAA. GAN and AAN. AUG. and AAA. which is not the canonical representative of any amino acid. For this reason. under the affine transformation t6 ◦ F . CAA. The other five amino acids. are Gln. we deleted the vertex with label UCC and its four adjacent edges in the last graph diagram. whose canonical triplets are. and Trp. It codes for Ser. GAA. Glu. the addition of the triplet CCG. As before. Amongst the 16 triplets which are labels of the vertexes. respectively. GAA. which are not represented in the hypercube NNC . 8 Plot of the general encoding function E in its two main forms. but this amino acid is already represented by the triplet AGC. under the affine transformation t56 ◦ F . these two triplets are the unitary images of the faces AU N and UGN. that is. In Fig. 4. belong to the hypercube NN A. that is. there is only one. 9. which is the image of NNC under the translation t56 .

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

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

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

a line parallel to a plane. they are parallel. The concept of n-dimensional hypercube Let us consider a vector space of the form V = Kn . that is. In this case. . the affine subspaces that contain the null vector. the identity v + W = W is obtained. From standard courses of linear algebra.. Concepts of point. K being a field. for a fixed linear subspace W. Consequently. where K is a field. . 0. etc. provided of a coordinate system. the set or family of all the affine subspaces or linear varieties contained in V . . that is. such that u − v ∈ W. line. if n is its dimension. 1.2. A. The points. The associated geometry to a vector space. . 0). every n-dimensional K-vector space is isomorphic to the space Kn . is the solution set of a linear system . We say that 2 affine subspaces v + W and u + U are parallel if W is a subspace of U . where the xi are elements of K. 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. they form a base of the vector spaces. respectively. Definition A. if the affine subspaces have the same associated subspaces. .236 Bulletin of Mathematical Biology (2007) 69: 215–243 difference vector u − v belongs to W. .2. The only affine subspace of dimension n is the whole space V . . in an n-dimensional vector space V . We call the associated geometry of a K-vector space V . . . . It is easy to show that these vectors are linearly independent and that they generate the space. The concept of parallelism Definition A. or as some vector which belongs to W. . 0. are the equivalent classes of the equivalence relation RW . Then we say that the space has been coordinatized. is the ordered set of vectors e1 = (1. . are generically called hyperplanes of the geometry. are called the coordinates of the vector. having the same dimension. 2. . also called points of the associated geometry. lines and planes in a geometry are. parallel planes. The other affine subspaces.3. . Hence.1. defined as the set of all the ordered pairs (u. . The isomorphism may be defined by the matching of any of the bases of V with the canonical bases of Kn . the affine subspaces of dimension 0.2. 1. the vectors are represented as n-tuples (x1 . As it is well known.2. The elements xi . If we take v as the null vector 0. 0). . all the vector subspaces are also affine subspaces. x2 . .3. it is known that every affine subspace v + W. Definition A. A. parallel cubes. in fact. In particular if W = U . The canonical base of Kn . or U is a subspace of W.2. xn ). and plane in a geometry. Thus. en = (0. all the unitary sets are affine subspaces of dimension zero. that is. and 2. The linear varieties or affine subspaces. e2 = (0. 0. . . n}. the k-dimensional for 2 < k < n. . . According to this definition we can talk of parallel lines. . 1) .. 0. v ) ∈ V × V . for i ∈ {1.

the six-dimensional hypercube of the 64 codons is illustrated. the solution set of the associated homogeneous system. Definition A. In Fig. is the so-called reduction module p.4. coincide with the extremes of the vectors e1 . 1973).2.. e2 and e3 of the canonical bases. The concept of coordinated affine subspaces Definition A. It is also called an orthotope because the angle of two adjacent edges is a right angle. A. the vector space Kn is called n-dimensional hypercube. where r denotes the rank of the associated matrix. which are remainders of the entire division by p. In this case. where (Z p ) denotes the Galois Field of p elements. In the case of the binary field (Z2 ). 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. i. . A coordinated affine subspace of the hypercube is called a hyperface of the hypercube (Z2 )6 . one of whose vertexes is the null vector 0 = (0.1. In fact. and all the edges have length 1. 5 of the main text.Bulletin of Mathematical Biology (2007) 69: 215–243 237 of n equations with n unknowns.3. In the particular case where K is the binary field Z2 = {0. 1}. The two-dimensional hyperfaces are simply the faces of the hypercube.1. where the eight triplets of zeros and ones.5. The zero-dimensional hyperfaces are the vertexes of the hypercube whereas the one-dimensional hyperfaces are the edges of it. x2 . the field of non-negative integers. Definition A.e. the norm is simply the number of ones that appear in the n-tuple. the 2 2 2 ordinary sum |v | = x1 + x2 + · · · + xn . xn ). its defect remainder in the entire division by p.4. Definition A. and the vertexes which are adjacent to it. We will call norm or length of a vector v = (x1 . A.1. being p a prime number. The ndimensional hypercube (Z2 )n is a regular polytope (Coxeter. 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 ..4. The dimension of W is equal to n − r . being the linear subspace.5. of two elements. The name hypercube comes from analogy with the three-dimensional case. a coordinated affine subspace is the solution set of a system whose associated matrix is diagonal. . the system with the same left parts and the right members equal to zero. 0). . it is also called the weight of the vector. The vectorial coordinated lines are usually called coordinated axes and the vectorial coordinated planes simply coordinated planes. which is always a nonnegative integer. i. . The assignment to every integer.e. . 0. represent the points which are the vertexes of a cube or regular hexahedron.

. y2 . It can be proved that the inner product and the length or norm are related in the following way. Transformations that preserve the distance Definition A. The concept of distance in the spaces (Z p )n . i. v ) of arbitrary vectors u and v . called permutation matrices. An isometric linear transformation will be called a linear isometry of the vector space. the following equality holds: |u + v | = |u| + |v | + 2 u.6. . A. A transformation f of the vector space V = (Z2 )n in itself. visualizing the space as a graph. every linear isometry performs a permutation on the set {e1 . . In vector spaces of the type (Z2 )n . x2 . . . 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. Actually. en . . v . the Hamming distance is the number of places in which both vectors differ. the linear isometries will also be called permutation transforms. and the affine coordinated planes are called faces of the hypercube.2.6. . and geometrically. . is called isometric if it preserves the distance between every two elements. . en }. For this reason. if d( f (u).e. Note that the norm or length of a vector coincides with the scalar product of the vector with itself.. v ) for every pair (u. w >= x1 y1 + x2 y2 + · · · + xn yn . this distance is the so-called Hamming distance. Definition A. A. It is easy to show that a linear isometry preserves the length or norm of every element of the space. We call scalar or inner product of the vectors v = (x1 . .1. . e2 . the affine coordinated subspaces are called hyperfaces of the hypercube. For every pair (u. In general. v ) ∈ V × V .238 Bulletin of Mathematical Biology (2007) 69: 215–243 In the more general case of a vector space (Z p )n . the affine coordinated lines are usually called edges. yn ) the integer < v. f (v )) = d(u. We define the distance between the vectors v and w as the norm of the difference vector v − w . the so-called hypercubes. xn ) and w = ( y1 . e2 .. being the subtraction the inverse operation of the addition in the vector space (Z p )n . 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 . . .2. have the property that the inverse of any of them is equal to its transposed. .6. it means the minimal number of edges between the two vertexes represented by the n-tuples. This kind of matrices. . widely used in coding theory and in criptology. In the case of the hypercube (Z2 )n .7. we can define the concept of scalar or inner product in the following way: Definition A.5.

one to each other. given by the function: v → tv . where i . We call affine transformation to every composed function of the form tv ◦ F . ti jk as the translation associated to the vector ei + e j + ek. A. from V to V . a linear isometry preserves the orthogonality of any pair of orthogonal vectors of the vector space.8.9. being the isomorphism. It can be proved that a linear isometry preserves not only the distance between vectors.1. Special notation. i. where F is a linear transformation or linear endomorphism of the vector space and tv is the associated translation of a fixed vector v . Definition A.9. A. a multiple rotation in the vector space. isomorphic to the additive group of V .. . if their scalar product is equal to zero. The concept of translation and affine transformations in a vector space Definition A. and only if. Definition A. We say that two vectors v and w of the space (Z p )n are orthogonal or perpendicular. j . k ∈ {1.9. T is. According to this definition the vectors of the canonical bases. The concept of orthogonality in a vector space (Z p )n Definition A. It means that the set T of all the translations of the vector space V is a group. and only in this case. 2. . Hence. .1. an abelian group which is a subgroup of the symmetric group S(V ) of all the bijections of the set V with itself. It is easy to prove that every affine transformation carries a linear variety v + W onto another linear variety. associated to the vectors v and w .e. If the affine transformation is bijective. but also the scalar product of any two vectors of the vector space.8. we call a translation associated to the vector v . For translations tv and tw .Bulletin of Mathematical Biology (2007) 69: 215–243 239 As every permutation is a composition of pairwise disjoint cycles. its linear component F is also bijective. every linear isometry can be interpreted as a composition of local rotations. This kind of base is usually called orthonormal base.1. additionally. It is clear that an affine transformation tv ◦ F is bijective. in fact. one to one. as ti j . a multiple rotation in (Z p )n can also be interpreted as a multiple rotation in Rn . As the set (Z p )n can be viewed as a subset of the Rvector space Rn . we will denote as ti the translation associated to the vector ei of the canonical base. the translation associated to the vector ei + e j . In the hypercube (Z2 )n . in any vector space of the form (Z p )n are orthogonal.2. if. to the bijective function tv : u → u + v . the composition tv ◦ tw is the translation tv+w . that is. the dimension is preserved. associated to the sum v + w of both vectors. being R the field of real numbers. Given a vector v of a vector space V . The linear isometries of the vector space (Z p )n will also be called permutation transforms or multiple rotations. and so on. n}. whose dimension is less than or equal to that of v + W.7. of length 1. from V onto T . . and they are. .

0. z. and G= 11. 0. the nucleotide A is replaced by the nucleotide G. Some biological concepts and their mathematical representation A. that every triplet XYZ is represented as a sextuple (x . to a distance h greater than 1. it is a transition. According to our representation of triplets as elements of the vector space (Z2 )6 . that is. The nucleotides A and G are called purines. It can be proved that the union of the k-dimensional hyperfaces v + W and ei + v + W. since A and G are both purines. we have assigned to every nucleotide a numerical pair in the following way: C= 00. 1. associated to it. purine to pyrimidine or pyrimidine to purine. 0. is called a right hyperprism of height h. The union of the hyperfaces v + W and u + v + W. a (k + 1)-dimensional hyperface is obtained. that is. A codon mutation is called a transition when it changes a base to a base of the same class. 0. by substitution of the base A by the base G. then the purine A is changed to the pyrimidine U. Then. if the triplet CUA is changed for the triplet CUG. y. 0. v ) of zeros and ones.10. 0. by the translation t5 . that is. it is a transversion.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 . Mutations Definition A.10. 0. but it is not a hyperface of the hypercube. 1. A right hyperprism is obtained from a hyperface by the adjunction of a translation of it.1. 1) and this transformation may be represented by the addition of the vector e5 = (0. For instance. A right hyperprism of height h. U= 10.9. U. G} is called a mutation. is an affine subspace. If the translation is to a distance equal to 1.3.2. that is. Definition A. Then. u. then the set u + v + W is also a k-dimensional hyperface. t . 1. Definition A. purine to purine or pyrimidine to pyrimidine. if the triplet CUA changes to the triplet CUG. A= 01. . when ei does not belong to W. We say that the mutation is simple if the substitution involves only one base of the triplet. 1. if CUA is changed to CUU. The substitution in a triplet of one or more of its three nucleotide bases within the set {C. 0). it means that the vector (0. On the the other hand. 0. Transitions and transversions. as an element of the hypercube (Z2 )6 . In our work. It follows. is a (k + 1)-dimensional hyperface. In this particular case the mutation is simple. and it is called a transversion when the change is from one class to another.10. and the nucleotides C and U are called pyrimidines. every mutation may be represented by means of a translation tv . when u is orthogonal to W and has length h greater than 1. This terminology is due to their chemical composition. 1) is changed to the vector (0. with h greater than 1. associated to some vector v . since it consists in the change of only one base. for instance. A.

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

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

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