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).

we used the following ordered assignment of the nucleotide bases: C↔ 00. (A. We also derive the general encoding function of the SGC. when the machinery of protein synthesis was imprecise. We hypothesize that in order to allow further evolution of the RNA genetic code. the constraint of having an intact message in only one reading frame has to be relaxed. 10. Jimenez-Monta no ´ et al.. For the interested reader.g.. and it represents the SGC code itself. A.3) associated to the RNA-less code can be inserted as pairwise disjoint four-dimensional affine subspaces in the six-dimensional hypercube (Coxeter. A3. 3.1. which also represent the integers 0. G). The article is organized as follows. In previous works (Sanchez et al. 2004a. 11 to the letters C. Given that C and G. Hence. the three four-dimensional hypercubes and the right hyperprism (A9. it is convenient to select the ordering in such a way that 00 is complementary of 11. 1. 1996. G). . G. A↔ 01. observing that 64 is equal not only to 43 but also to 26 . and 01 is to 10.b. purines A and G are represented by the odd numbers 1 and 3. suggested to organize the codon table as a six-dimensional hypercube ´ or six-dimensional vector space over the binary field. 1968. A. Interestingly. 1973. A and U are complementary to each other in double stranded DNA. in the binary numerical system. 01. we have only eight possible selections: (C. 5). Therefore. With this ordering. we discuss our findings in terms of the origin and evolution of the SGC. U. Sanchez et al. This encoding function adopts different forms for different subsets of codons. 1977. 2. 1995) with the SGC. Konecny et al. Theoretical background The standard table of codon assignments derives from the obvious representa´ ˜ tion of the triplet code as a 4 × 4 × 4 cube. C.1) as derived by concepts of combinatorial geometry. about the concepts of algebra and geometry is provided at the end of this article. associated to the Extended RNA code. G. second and third reading frames. and. Several authors (e. 1. U). G. First. U↔ 10. an Appendix which is referred to throughout this work.. we search for symmetries and patterns in both the SGC and the RNA code. 2005). U. may be done in 24 = 4! ways. given that translational and transcriptional errors were probably of great importance early in the history of life. we consider not a strict comma-less code as proposed by Crick et al. This assignment of the duplets 00. To this end. Here.Bulletin of Mathematical Biology (2007) 69: 215–243 217 To our knowledge. (A. Finally.1. Eigen and Schuster. A1. there has not been systematic studies that relate the RNA code (Crick. we define an encoding function for each of the three four-dimensional hypercubes. (C. 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. U). U. two algebras were presented to reflect the relationship between codon assigment and the physicochemical aspects of the amino acids. we show that each reading frame can be represented as a four-dimensional hypercube (A3. whereas the pyrimidines C and U are represented by the even numbers 0 and 2. 2005). G↔ 11.1. A.. C. and an encoding function for the RNA-less code. (1957) but rather an RNA code which can be translated in the first. Fig. and we show that this function is an integration of the different encoding functions of the above-mentioned four sets. Next.

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

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

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

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

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

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

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

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

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

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

ACU → ACC AUC. t16 and t56 the composed translations t1 ◦ t6 and t5 ◦ t6 . which is a three-dimensional hyperface of the hypercube NYR. GAU → GAC GGC. The explicit definition of the function F1 is: F1 ACC. specified by the F RF .3. UCG → UC A CC A. F2 coincides with the restriction to NYR of the affine transformation t6 ◦ F . AUU → AUC GCC. which assigns to every set of triplets. AGU → AGC GAC. it is so. belong to the cube NYA. 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 . In most cases. Actually. GGU → GGC (for Threonine) (for Isoleucine) (for Alanine) (for Valine) (for Asparagine) (for Serine) (for Aspartic acid) (for Glycine) 3.4. for 13 out of the 16 triplets. we define a function F2 . we define a function F3 . The encoding function for the triplets of the second reading frame In the hypercube of codons of the form NYR. with the only exception of AUG. associated to the SRF of the Extended RNA code. described above. For UUG and UUA. GCG → GC A AC A. UUA and AUG. whereas for AUG it acts as the composition t56 ◦ F . UUG → CU A UC A.228 Bulletin of Mathematical Biology (2007) 69: 215–243 and NNC . The explicit definition of the function F2 is: F2 CU A. encoding for the same amino acid. GUG → GU A AUG → AUG AU A → AU A GC A. The function F2 is related to the linear transformation F . F2 behaves as the composition t16 ◦ F . UU A. CUG. respectively. But the function F2 is not exactly a linear projection of NYR onto NYA. according to the linear order in the whole set of triplets. 3. which assigns to every set of triplets. in the following way. CCG → CC A GU A. the minor. AAU → AAC AGC. GCU → GCC GUC. . as is the case of F1 in the hypercube RNY. Here. we are taking the cube RNC as a canonical representation of the set of eight amino acids. the composition of F with the translation t6 . associated to the TRF of the Extended RNA code. with only three exceptions: UUG. The encoding function for the triplets of the third reading frame In the hypercube of codons of the form YRN. GUU → GUC AAC.

UAA. belong to the cube YRC . CCU → CCC CUC. UUU → UUC AAA. CAG and CAA. For UGG. the function F3 coincides with the restriction to YRN of the linear transformation F : XYZ → XYC . encoding for the same amino acid. UCC. 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. A graphical representation of the encoding functions F1 . However. CCC. F2 . For UAG. GAG. note that several codons have only . 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 . U AU → U AC C AC. However. AGG → AGA GAA. In order to build this plot. F4 acts as the restriction to RRR of the affine transformation t6 ◦ F . UUU. which assigns to every set of triplets. associated to the RNY-less code we can define a function F4 . GGG. F3 coincides with the restriction to YRN of the affine transformation t6 ◦ F For UGA. UCU. for 10 triplets. U AG. GGA. 7.5. the function F4 acts as the restriction to the set YYY of the linear transformation F . The linearity of the encoding functions is apparent even considering few departures. AGA. 3. and F4 is shown in Fig. GAG → GAA GGA. C AU → C AC UGG → UGG U AA. according to the linear order in the whole set of triplets. the minor. 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. AAA. CUU. F3 coincides with the restriction to YRN of the affine transformation t36 ◦ F . AAG → AAA AGA. GAA. UGA → U AA C AA. UGU → UGC CGC. for the triplets AAG. CUC. CGA. we have considered the lexicographic order of the triplets used above and we have assigned to this order the corresponding integer values from 0 to 63. The explicit definition of the function F3 is: F3 UGC. the minor. The encoding function for the triplets of the RNA-less code In the vector subspace of the codons of the form RRR or YYY. F3 . AGG. according to the linear order in the whole set of triplets. In fact. F3 coincides with the restriction to YRN of the affine transformation t56 ◦ F .Bulletin of Mathematical Biology (2007) 69: 215–243 229 encoding for the same amino acid. The explicit definition of the function F4 is: F4 CCC. CGG → CGC U AC. UCU → UCC UUC. CGU. UUC.

F3 and. F2 and F4 . hence these codons have two or three representative codons. one representative codon but some of them appear in two (e. The general encoding function for amino acids and stop codons in the hypercube of 64 triplets As it is well known. In Table 3. 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. as well as a combinatorial geometry. 1}.6. we derive an algebraic function. F4 . also called six-dimensional hypercube. It is done by means of the assignment C ↔ 00.g. which leads to the representation of each triplet as a sextuple of zeros and ones. the triplets .g. F3 and F4 . Ser appears when F1 . F2 . associated to the vectorial structure. G ↔ 11.230 Bulletin of Mathematical Biology (2007) 69: 215–243 Encoding Function F1. F2. are applied) of the subsets. Herein. We have defined in the set of 64 triplets a structure of a vector space over the binary field Z2 = {0. U ↔ 10. This correspondence involves the addition operation and the vectorial algebraic structure in the set of triplets. 3. that is. A ↔ 01. as a vector of the six-dimensional vector space (Z2 )6 . 7 Local encoding functions F1 . 61 out of the 64 triplets code for the 20 primary amino acids that are the building blocks of proteins. A graphical representation of the local encoding functions F1 . Then a given amino acid will appear at different ordinate values. F2 . which assigns to every triplet its associated amino acid or its termination mark. Pro appears when F2 and F4 are applied) or even three (e.

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

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

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

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

Bulletin of Mathematical Biology (2007) 69: 215–243 235 acids of the RNY are also coded by triplets that appear in the SRF. The dimension of a linear variety is defined as the dimension of its associated vector subspace. or linear variety contained in V . ´ 1999. and with that of vector subspace of a vector space. respectively. Concept of affine space and its dimension Definition A. for a fixed subspace W. A. Appendix A. also called a K-vector space. Szathmary. Given the uneven degeneracy of the genetic code it is appealing that the general encoding function is almost linear. 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.1.2. It is also worth to mention that in present day DNA virus such as vaccinia virus. Remarks. The set v + W is also called a coset or adjoint class of the subgroup W. a single messenger RNA (mRNA) is able to encode three different proteins because messages contain three distinct putative translation initiation sites.1. to any of the three four-dimensional hypercubes of the Extended RNA code. and it contains the element v . to every subset of the form v + 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. Konecny et al. we can say that considering the symmetries of both the Extended RNA code and the RNA-less code. The phenotypic graph of amino acids is also a novel finding whose image resembles. 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. and there are new triplets which altogether specify nine new amino acids. In the vaccinia virus. To our knowledge. 1990). 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.. For a vector space V we call affine subspace of V . The associated vector subspace W. the primaeval RNY code was already frozen and that it evolved like a replicating and growing icicle. for a linear variety.. Definition A. where W is a vector subspace and v is a fixed vector.1. genes are transcribed in a frame-shift fashion (Keck. The subspace W is called the associated vector subspace of the linear variety v + W. 1990). et al. is unique.1. it means that their . In other words. Mathematical and biological background We assume that the reader is familiar with the concept of a vector space over a scalar field K.. In the context of the frozen concept. in part. but the vector v may be any of the elements of the set v + W. When two vectors u and v define the same linear variety. We recall that every vector space is an abelian group for the addition operation. as well as with the concept of a linear transformation or linear endomorphism of a vector space.

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

coincide with the extremes of the vectors e1 .4. which are remainders of the entire division by p. A coordinated affine subspace of the hypercube is called a hyperface of the hypercube (Z2 )6 . where (Z p ) denotes the Galois Field of p elements. 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 . The concept of coordinated affine subspaces Definition A. The vectorial coordinated lines are usually called coordinated axes and the vectorial coordinated planes simply coordinated planes. The zero-dimensional hyperfaces are the vertexes of the hypercube whereas the one-dimensional hyperfaces are the edges of it.. the field of non-negative integers. the 2 2 2 ordinary sum |v | = x1 + x2 + · · · + xn . a coordinated affine subspace is the solution set of a system whose associated matrix is diagonal. which is always a nonnegative integer.4. . the six-dimensional hypercube of the 64 codons is illustrated. A. In the particular case where K is the binary field Z2 = {0. The dimension of W is equal to n − r . the vector space Kn is called n-dimensional hypercube.5. Definition A. e2 and e3 of the canonical bases. x2 .1. The ndimensional hypercube (Z2 )n is a regular polytope (Coxeter.5. is the so-called reduction module p.. the system with the same left parts and the right members equal to zero. of two elements. and all the edges have length 1. 5 of the main text. The name hypercube comes from analogy with the three-dimensional case. and the vertexes which are adjacent to it. . In this case.1.2. its defect remainder in the entire division by p. 0). The assignment to every integer. where the eight triplets of zeros and ones.4. xn ). In Fig. represent the points which are the vertexes of a cube or regular hexahedron. i. We will call norm or length of a vector v = (x1 . In the case of the binary field (Z2 ). the solution set of the associated homogeneous system. 0.e.1. In fact. The two-dimensional hyperfaces are simply the faces of the hypercube. being the linear subspace. the norm is simply the number of ones that appear in the n-tuple. it is also called the weight of the vector. . 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. being p a prime number. 1973). where r denotes the rank of the associated matrix.Bulletin of Mathematical Biology (2007) 69: 215–243 237 of n equations with n unknowns. .3. Definition A. A. Definition A. i. .e. one of whose vertexes is the null vector 0 = (0. 1}. It is also called an orthotope because the angle of two adjacent edges is a right angle.

. In the case of the hypercube (Z2 )n . For every pair (u. It is easy to show that a linear isometry preserves the length or norm of every element of the space.7.2.. . the affine coordinated lines are usually called edges. widely used in coding theory and in criptology. and the affine coordinated planes are called faces of the hypercube.6. . the so-called hypercubes. Definition A. v ) ∈ V × V . A. . The concept of distance in the spaces (Z p )n . the following equality holds: |u + v | = |u| + |v | + 2 u. and geometrically. For this reason. is called isometric if it preserves the distance between every two elements. en . Transformations that preserve the distance Definition A. .. Actually.2. have the property that the inverse of any of them is equal to its transposed.1. 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. e2 . . A. The concept of permutation transform or multiple rotation in the vector space V = (Z p )n As the only vectors of length 1 in the vector space are the canonical vectors e1 . . An isometric linear transformation will be called a linear isometry of the vector space.5. We define the distance between the vectors v and w as the norm of the difference vector v − w . the linear isometries will also be called permutation transforms. xn ) and w = ( y1 . x2 . if d( f (u). In general. this distance is the so-called Hamming distance. visualizing the space as a graph. This kind of matrices. y2 . We call scalar or inner product of the vectors v = (x1 . . called permutation matrices. being the subtraction the inverse operation of the addition in the vector space (Z p )n . . it means the minimal number of edges between the two vertexes represented by the n-tuples. we can define the concept of scalar or inner product in the following way: Definition A. . e2 . . the affine coordinated subspaces are called hyperfaces of the hypercube. i. every linear isometry performs a permutation on the set {e1 . v . It can be proved that the inner product and the length or norm are related in the following way.6. A transformation f of the vector space V = (Z2 )n in itself. . .6. w >= x1 y1 + x2 y2 + · · · + xn yn . en }. f (v )) = d(u. v ) for every pair (u. yn ) the integer < v. v ) of arbitrary vectors u and v .238 Bulletin of Mathematical Biology (2007) 69: 215–243 In the more general case of a vector space (Z p )n .e. Note that the norm or length of a vector coincides with the scalar product of the vector with itself. . . the Hamming distance is the number of places in which both vectors differ. In vector spaces of the type (Z2 )n .

j . an abelian group which is a subgroup of the symmetric group S(V ) of all the bijections of the set V with itself. According to this definition the vectors of the canonical bases.. If the affine transformation is bijective.9. a multiple rotation in (Z p )n can also be interpreted as a multiple rotation in Rn . from V to V . n}.2. associated to the vectors v and w . . one to each other. one to one. and so on. The concept of translation and affine transformations in a vector space Definition A. whose dimension is less than or equal to that of v + W. Special notation. k ∈ {1. Definition A. The linear isometries of the vector space (Z p )n will also be called permutation transforms or multiple rotations. but also the scalar product of any two vectors of the vector space. As the set (Z p )n can be viewed as a subset of the Rvector space Rn .9. a multiple rotation in the vector space. i. from V onto T . to the bijective function tv : u → u + v . In the hypercube (Z2 )n . and only if. Given a vector v of a vector space V . It is easy to prove that every affine transformation carries a linear variety v + W onto another linear variety.e. Hence.Bulletin of Mathematical Biology (2007) 69: 215–243 239 As every permutation is a composition of pairwise disjoint cycles. the dimension is preserved. being the isomorphism.1. isomorphic to the additive group of V . the translation associated to the vector ei + e j . A. in any vector space of the form (Z p )n are orthogonal. 2.9. It is clear that an affine transformation tv ◦ F is bijective. It means that the set T of all the translations of the vector space V is a group. every linear isometry can be interpreted as a composition of local rotations. its linear component F is also bijective. . This kind of base is usually called orthonormal base.1.1. and only in this case.8. We call affine transformation to every composed function of the form tv ◦ F . and they are. we call a translation associated to the vector v .8. associated to the sum v + w of both vectors. of length 1. that is. the composition tv ◦ tw is the translation tv+w . T is. a linear isometry preserves the orthogonality of any pair of orthogonal vectors of the vector space. ti jk as the translation associated to the vector ei + e j + ek. if their scalar product is equal to zero. as ti j . It can be proved that a linear isometry preserves not only the distance between vectors. . A. For translations tv and tw . where F is a linear transformation or linear endomorphism of the vector space and tv is the associated translation of a fixed vector v . . We say that two vectors v and w of the space (Z p )n are orthogonal or perpendicular. given by the function: v → tv . where i .7. we will denote as ti the translation associated to the vector ei of the canonical base. additionally. in fact. being R the field of real numbers. The concept of orthogonality in a vector space (Z p )n Definition A. Definition A. . if.

z. that is. 1. by the translation t5 . if CUA is changed to CUU. that is. If the translation is to a distance equal to 1. 0). if the triplet CUA is changed for the triplet CUG. Mutations Definition A. it is a transversion. Definition A. Definition A. Then. The union of the hyperfaces v + W and u + v + W. every mutation may be represented by means of a translation tv . that is. 0. purine to purine or pyrimidine to pyrimidine. For instance. U. it means that the vector (0. t . for instance. On the the other hand. A right hyperprism is obtained from a hyperface by the adjunction of a translation of it. 0. In our work.2. 0.1. purine to pyrimidine or pyrimidine to purine. 1. 1) and this transformation may be represented by the addition of the vector e5 = (0. the nucleotide A is replaced by the nucleotide G.10. as an element of the hypercube (Z2 )6 . It follows. 0. y.10. if the triplet CUA changes to the triplet CUG. In this particular case the mutation is simple. According to our representation of triplets as elements of the vector space (Z2 )6 . when u is orthogonal to W and has length h greater than 1. 0. 1. A codon mutation is called a transition when it changes a base to a base of the same class. A= 01. we have assigned to every nucleotide a numerical pair in the following way: C= 00. is a (k + 1)-dimensional hyperface. a (k + 1)-dimensional hyperface is obtained. We say that the mutation is simple if the substitution involves only one base of the triplet. then the set u + v + W is also a k-dimensional hyperface. that every triplet XYZ is represented as a sextuple (x . u. . v ) of zeros and ones.3. Transitions and transversions. A right hyperprism of height h. by substitution of the base A by the base G. G} is called a mutation. This terminology is due to their chemical composition. it is a transition. A. when ei does not belong to W. 0. 1. 0.240 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 . that is. Then. then the purine A is changed to the pyrimidine U. U= 10. and it is called a transversion when the change is from one class to another. to a distance h greater than 1. The substitution in a triplet of one or more of its three nucleotide bases within the set {C. but it is not a hyperface of the hypercube. is called a right hyperprism of height h. 0. Some biological concepts and their mathematical representation A.9. 1) is changed to the vector (0. and the nucleotides C and U are called pyrimidines. is an affine subspace. 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. with h greater than 1. The nucleotides A and G are called purines.10. and G= 11. associated to it. since it consists in the change of only one base. associated to some vector v .

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

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

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

Sign up to vote on this title
UsefulNot useful