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

1). U.pyrimidine. U. 1}(A3. G. The same happens with some of the angles of a three-dimensional cube. A).any nucleotide). In the vector space structure of the hypercube the addition is the so-called sum module 2. 1986) which comprises the codons with an RNY pattern (R . but in our pictures many of them look like acute or obtuse. but it is possible to show that the remaining ones lead to the same results. The RNA code and its three reading frames: The extended code A primaeval RNA World was proposed (Gilbert. G. This sum. we have selected the first. 2. Since the primitive translational apparatus may have been imperfect. N .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. C. C). The vector space (Z2 )6 is an orthotope. all the angles between adjacent edges are supposed to be right angles. (U. This is so because the picture is a projection of a six-dimensional figure over a plane. which is a vector space over the binary field Z2 = {0. From these orderings. A). the so-called Klein Four Group. shifts in the reading frame probably occurred. and codons with an NYR and YRN patterns emerged. for the elements of the field Z2 . according to the terminology of Coxeter (1973) (A3). is defined in Table 1.1. 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 . which is the operation table of a known abelian group of order 4. A. When Table 1 is translated to the nucleotide bases it gives Table 2. and (G. In the hypercube. widely used in symbolic logic. Y . C. Results 2. as in every orthotope. when they are portrayed in a plane. This group is isomorphic (A3) to the group of all the symmetries of a plane rectangle (not including the square). C). The 64 sextuples of zeros and ones generate the so-called six-dimensional hypercube. These three patterns altogether are here defined as the codons of the Extended RNA code.purine. A. bit by bit. (G. also called XOR operation.

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

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

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

NYR and YRN. second and third reading frames.222 Bulletin of Mathematical Biology (2007) 69: 215–243 from the other by means of the translation t2 .4. e2 → e6 . We will consider the union of the sets of codons of the form RNY. as an extension of the primaeval RNY code. Thirteen out of the 16 triplets of the form YRN code for six new amino acids. or four-dimensional hyperface of the six-dimensional hypercube of the 64 triplets (Fig. the addition of the vector e4 + e6 which involves transversions in the second and third nucleotides. The Hamming distance between a vertex and its image. . 2c). under the translation t6 . that is. e4 → e2 . 3a) and YRR (Fig. e5 → e3 . and their correspondence with 17 amino acids and stop codons. or double rotation. The matrix of this linear transformation is orthogonal with determinant 1. is equal to 1. which converts RNY into NYR. The third reading frame If in a sequence of triplets of the form RNY. by the addition of the codon ACC to every triplet. This function maps every triplet XYZ onto the triplet YZX. which is a coordinated affine subspace. The set of triplets of the form YRN is a disjoint set with the sets RNY and NYR. The hypercube YRN is the image of the NYR under the non-singular linear transformation e1 → e5 . associated to the canonical vector e2 . 2. NYR and YRN. Then. the addition of the codon CCA to every triplet. and the other three codons correspond to a termination signal. then triplets of the form YRN will be translated. The other three codons are the so-called stop codons. which complete a set of 17 out of the 20 primary amino acids. Each cube is obtained from the other by the translation t6 . The hypercube NYR is the image of the RNY under the non-singular linear transformation e1 → e5 . which is a coordinated affine subspace. that is. associated to the canonical vector e6 . which is the same basis permutation. e2 → e6 . without repetitions of any of those found in the FRF or the SRF. e6 → e4 . e5 → e3 . 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. due to a slippage. 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. comprises the Extended RNA code. In summary. e4 → e2 . ignoring the first (R) and the second (N) nucleotides. of the vector e4 that represents a transversion in the second nucleotide. YRY is obtained from YYY by addition of the codon CAC. where 45 out of them specify 17 of the primary amino acids. YRR is obtained from YYY by the addition of the triplet CAA. e3 → e1 . e6 → e4 . the triplets of the form YRN give rise to six new amino acids. those associated to the first. that is. which is defined by an even permutation of the canonical base. the set of 48 codons of the form RNY. that is. It means that the set of codons of the form NYR constitutes also a four-dimensional hypercube. 3b). and it can also be interpreted as a double rotation in the six-dimensional Euclidean vector space R6 (A7). the reading starts at the third nucleotide Y. It means that the set of codons of the form YRN constitutes also a four-dimensional hypercube. The set of codons of the form YRN is the union of the cubes YRY (Fig. under the translation t2 . 3c). is equal to 1. The Hamming distance between a vertex and its image. e3 → e1 . or a four-dimensional hyperface of the six-dimensional hypercube of the 64 triplets (Fig. that is.

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

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

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

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

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

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

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

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

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

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

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

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

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

Then we say that the space has been coordinatized. the k-dimensional for 2 < k < n.2. is the ordered set of vectors e1 = (1. parallel planes. 0.1. 0). . Thus. for i ∈ {1. where K is a field. . The canonical base of Kn . . Hence. The only affine subspace of dimension n is the whole space V . also called points of the associated geometry. 0. respectively. the affine subspaces that contain the null vector. e2 = (0. . The points. v ) ∈ V × V . 1. The isomorphism may be defined by the matching of any of the bases of V with the canonical bases of Kn . In particular if W = U . From standard courses of linear algebra. if the affine subspaces have the same associated subspaces.2. Definition A. . the affine subspaces of dimension 0. A. The associated geometry to a vector space.236 Bulletin of Mathematical Biology (2007) 69: 215–243 difference vector u − v belongs to W. the set or family of all the affine subspaces or linear varieties contained in V . in an n-dimensional vector space V . that is. . 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. 1. they are parallel.3. K being a field. . defined as the set of all the ordered pairs (u. Definition A. We call the associated geometry of a K-vector space V . having the same dimension. n}.. . 0. are the equivalent classes of the equivalence relation RW . line. If we take v as the null vector 0. that is. in fact. where the xi are elements of K. xn ). every n-dimensional K-vector space is isomorphic to the space Kn . lines and planes in a geometry are. . the vectors are represented as n-tuples (x1 . and 2. or as some vector which belongs to W. We say that 2 affine subspaces v + W and u + U are parallel if W is a subspace of U . 2.2. The other affine subspaces. or U is a subspace of W. . are generically called hyperplanes of the geometry..2. and plane in a geometry. are called the coordinates of the vector. that is.3. . . en = (0. x2 . It is easy to show that these vectors are linearly independent and that they generate the space. The concept of parallelism Definition A. Consequently. According to this definition we can talk of parallel lines. such that u − v ∈ W. The elements xi . the identity v + W = W is obtained. A. As it is well known. all the unitary sets are affine subspaces of dimension zero. all the vector subspaces are also affine subspaces. it is known that every affine subspace v + W. . . parallel cubes. if n is its dimension. for a fixed linear subspace W. etc.2. they form a base of the vector spaces. The concept of n-dimensional hypercube Let us consider a vector space of the form V = Kn . . In this case. provided of a coordinate system. . . . is the solution set of a linear system . . 1) . . 0. . a line parallel to a plane. 0). Concepts of point. The linear varieties or affine subspaces. .

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

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

We say that two vectors v and w of the space (Z p )n are orthogonal or perpendicular. associated to the sum v + w of both vectors. It is clear that an affine transformation tv ◦ F is bijective. It can be proved that a linear isometry preserves not only the distance between vectors. . every linear isometry can be interpreted as a composition of local rotations. . isomorphic to the additive group of V . and they are. as ti j . we will denote as ti the translation associated to the vector ei of the canonical base. of length 1. It is easy to prove that every affine transformation carries a linear variety v + W onto another linear variety. For translations tv and tw . being R the field of real numbers. additionally. Definition A. If the affine transformation is bijective.9. A. one to one.Bulletin of Mathematical Biology (2007) 69: 215–243 239 As every permutation is a composition of pairwise disjoint cycles.e. one to each other. Definition A. n}. where i . As the set (Z p )n can be viewed as a subset of the Rvector space Rn . to the bijective function tv : u → u + v . The concept of translation and affine transformations in a vector space Definition A.. that is. The concept of orthogonality in a vector space (Z p )n Definition A. . the dimension is preserved. from V to V .9.1. a linear isometry preserves the orthogonality of any pair of orthogonal vectors of the vector space. a multiple rotation in (Z p )n can also be interpreted as a multiple rotation in Rn .2. It means that the set T of all the translations of the vector space V is a group.7. . The linear isometries of the vector space (Z p )n will also be called permutation transforms or multiple rotations. being the isomorphism. whose dimension is less than or equal to that of v + W. 2. This kind of base is usually called orthonormal base.9. T is.1. and so on. a multiple rotation in the vector space.8. given by the function: v → tv . According to this definition the vectors of the canonical bases. ti jk as the translation associated to the vector ei + e j + ek. In the hypercube (Z2 )n . we call a translation associated to the vector v . Special notation. but also the scalar product of any two vectors of the vector space. if their scalar product is equal to zero. associated to the vectors v and w . the translation associated to the vector ei + e j . k ∈ {1.8.1. and only if. from V onto T . A. and only in this case. in any vector space of the form (Z p )n are orthogonal. its linear component F is also bijective. if. Given a vector v of a vector space V . Hence. j . an abelian group which is a subgroup of the symmetric group S(V ) of all the bijections of the set V with itself. We call affine transformation to every composed function of the form tv ◦ F . in fact. the composition tv ◦ tw is the translation tv+w . where F is a linear transformation or linear endomorphism of the vector space and tv is the associated translation of a fixed vector v . i. .

but it is not a hyperface of the hypercube. 0. is a (k + 1)-dimensional hyperface. 0). In our work. then the set u + v + W is also a k-dimensional hyperface. that every triplet XYZ is represented as a sextuple (x . According to our representation of triplets as elements of the vector space (Z2 )6 . that is. 0. A. It can be proved that the union of the k-dimensional hyperfaces v + W and ei + v + W. 0. It follows. 1) and this transformation may be represented by the addition of the vector e5 = (0. Then. if the triplet CUA changes to the triplet CUG.2. that is. 1.3. that is. 0.10. Definition A. U= 10. A codon mutation is called a transition when it changes a base to a base of the same class. u. purine to pyrimidine or pyrimidine to purine. 0. U. it is a transition. associated to some vector v . we have assigned to every nucleotide a numerical pair in the following way: C= 00. and G= 11. every mutation may be represented by means of a translation tv . On the the other hand. if CUA is changed to CUU. A right hyperprism of height h. The nucleotides A and G are called purines. The union of the hyperfaces v + W and u + v + W. 1. for instance.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 . z. 1. is called a right hyperprism of height h. by the translation t5 . as an element of the hypercube (Z2 )6 . it is a transversion. when u is orthogonal to W and has length h greater than 1. We say that the mutation is simple if the substitution involves only one base of the triplet. v ) of zeros and ones. A= 01. This terminology is due to their chemical composition. Transitions and transversions. associated to it. If the translation is to a distance equal to 1. 1. 1) is changed to the vector (0.1. it means that the vector (0. A right hyperprism is obtained from a hyperface by the adjunction of a translation of it. that is. 0. purine to purine or pyrimidine to pyrimidine. and it is called a transversion when the change is from one class to another. since it consists in the change of only one base. to a distance h greater than 1. the nucleotide A is replaced by the nucleotide G. Some biological concepts and their mathematical representation A. 0. G} is called a mutation. Then. by substitution of the base A by the base G. a (k + 1)-dimensional hyperface is obtained.9. if the triplet CUA is changed for the triplet CUG. and the nucleotides C and U are called pyrimidines. Mutations Definition A. 0.10. is an affine subspace. when ei does not belong to W. t . with h greater than 1.10. y. For instance. Definition A. . The substitution in a triplet of one or more of its three nucleotide bases within the set {C. In this particular case the mutation is simple. since A and G are both purines. then the purine A is changed to the pyrimidine U.

the Hamming distance between a base and its complementary is always equal to 2. t3 . that is. are obtained. This triplet is usually called the complementary codon or the anticodon of XYZ. t13 . 1. the triplet GGG. of the six vectors of the canonical base. which have odd subindexes. Hence. defined by the vector M. The set of mutations which are transitions is closed under the composition. As T is an abelian group. and t6 . y. that is. t3 and t5 . All the other mutations are transversions or compositions of transversions with transitions.Bulletin of Mathematical Biology (2007) 69: 215–243 241 A. as result. which assigns to every triplet XYZ the triplet X Y Z . it means the addition of the triplet GGG to XYZ. Analogously. According to our numerical representation. For this reason. the subgroup of pure transitions is a normal subgroup of T . From the algebraic point of view. for they differ in two of their components. When U is changed by T (thymine). according to the addition operation illustrated in Table 2. The mentioned function is the translation t M . v ). Complementary bases For biological reasons. and Z. CUC and CCU. the addition to the vector (x . . t4 . respectively. t4 . Notice that the Hamming distance between a codon and its complementary is always equal to 6. are obtained by additions of the triplets ACC. The translation t M performs a transversion in each of the three components of the triplet. e3 and e5 . CAC and CCA (see Table 2). represented by the translations t1 . and t6 of even subindexes. and are represented by the numerical pairs 00 and 11. and Z are the complements of X. as additions of the triplets UCC.12. and the identity function may be considered as a transition. 1). one to each other. while the basic transversions are represented by the translations t2 . whereas the basic transversions represented by the translations t2 . t1 . which is the sum e1 + e2 + e3 + e4 + e5 + e6 . In the Boolean structure of the hypercube (Z2 )6 this transformation is the so-called Boolean negation. The sum of a triplet with its complementary gives. t35 and t135 being t0 the identity function. Y. the associated to the vector M. the set of all the transitions is a group of order 8.11. the purine A and the pyrimidine U are considered to be complement of each other. t15 . A. the vector M = (1. The basic transitions. u. the pyrimidine C and the purine G. the maximum of all the possible distances in the six-dimensional hypercube. Algebraic representations of transitions and transversions Notice that the basic transitions in the hypercube (Z2 )6 are represented by the translations t1 . associated to the triplet. they have been represented by complementary numerical pairs 10 and 01. Y . z. that is. are considered complementary bases. t3 . those associated to the canonical vectors e1 . t . 1. The eight transitions are represented by the translations: t0 . the base A is paired with T in the double helix structure of DNA. respectively. They also form a pair in the double helix structure. that is. and it consists in the substitution of every zero by ones and every one by zeros. In the vector space of the 64 triplets we can define a bijective function. 1. 1. subgroup of the group T of 64 possible mutations. and t5 . where the components X . t5 .

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

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