This action might not be possible to undo. Are you sure you want to continue?

# 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).

1996. This encoding function adopts different forms for different subsets of codons. Fig.1. we deﬁne an encoding function for each of the three four-dimensional hypercubes. Several authors (e. there has not been systematic studies that relate the RNA code (Crick. G↔ 11. 1. given that translational and transcriptional errors were probably of great importance early in the history of life. Here. we show that each reading frame can be represented as a four-dimensional hypercube (A3.3) associated to the RNA-less code can be inserted as pairwise disjoint four-dimensional afﬁne subspaces in the six-dimensional hypercube (Coxeter. First.. when the machinery of protein synthesis was imprecise. To this end. and an encoding function for the RNA-less code. This assignment of the duplets 00. A. (1957) but rather an RNA code which can be translated in the ﬁrst. and we show that this function is an integration of the different encoding functions of the above-mentioned four sets. Hence. We hypothesize that in order to allow further evolution of the RNA genetic code. suggested to organize the codon table as a six-dimensional hypercube ´ or six-dimensional vector space over the binary ﬁeld. about the concepts of algebra and geometry is provided at the end of this article. 2005). C. two algebras were presented to reﬂect the relationship between codon assigment and the physicochemical aspects of the amino acids. and it represents the SGC code itself. U↔ 10. Next. G. The article is organized as follows. 01.1. Sanchez et al. (C. the constraint of having an intact message in only one reading frame has to be relaxed. 1. Konecny et al. With this ordering. it is convenient to select the ordering in such a way that 00 is complementary of 11. A↔ 01. 2004a. we discuss our ﬁndings in terms of the origin and evolution of the SGC. Therefore. Jimenez-Monta no ´ et al. A and U are complementary to each other in double stranded DNA. 1977. in the binary numerical system. G. the three four-dimensional hypercubes and the right hyperprism (A9. Finally. C. associated to the Extended RNA code. U. A.g. we consider not a strict comma-less code as proposed by Crick et al. G). Interestingly. U). 2005). and. we have only eight possible selections: (C. second and third reading frames. U). 3. .b. U. observing that 64 is equal not only to 43 but also to 26 . We also derive the general encoding function of the SGC. A3. 1995) with the SGC. In previous works (Sanchez et al.1.. For the interested reader. Eigen and Schuster. purines A and G are represented by the odd numbers 1 and 3. A1. G). 5).. U. (A. an Appendix which is referred to throughout this work. A. 10. we used the following ordered assignment of the nucleotide bases: C↔ 00. whereas the pyrimidines C and U are represented by the even numbers 0 and 2. may be done in 24 = 4! ways. (A.Bulletin of Mathematical Biology (2007) 69: 215–243 217 To our knowledge. 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. G. and 01 is to 10.1) as derived by concepts of combinatorial geometry. Theoretical background The standard table of codon assignments derives from the obvious representa´ ˜ tion of the triplet code as a 4 × 4 × 4 cube.. 1973. 1968. Given that C and G. 2. 11 to the letters C. we search for symmetries and patterns in both the SGC and the RNA code. which also represent the integers 0.

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

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

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

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

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

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

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

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

u. It is the solution subspace of the homogeneous linear . UGU CAA. it is well known that most mutations in the third base does not change the corresponding amino acid (the wobble hypothesis) and therefore we started with a function that maps the third nucleotide of a codon to zero. the ﬁrst. UGA UGG AAA. A. The image of F . the triplets of the form CCZ.CCA. the representative of Ala is the triplet GCC. denoted by Im( F ). GAU GGC. Biologically. GUA. u. ACU.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. The endomorphism F belongs to a class of endomorphisms. CUG. GUU. UUU representations is the genetic code. ACA. AGA. The kernel of F is the two-dimensional subspace of vectors of the form (0. t . y. those that end with the nucleotide C. For example. which are called projections and are characterized by the property of idempotency. we consider the linear transformation. AAG GAA. GGG CCC. CUU. it changes the third nucleotide by the nucleotide C. a vectorial four-dimensional hypercube or a hyperface of the six-dimensional hypercube. AGC UAC. GUG AAC. of the vector space (Z2 )6 . CGU. CGG. the correspondence of the ordered set of triplets that specify every amino acid is illustrated. v ) → (x . 3. t . CAG UAA. ACG AUC. GAG UUC. the kernel of F is a face of the hypercube. CCG CUC. 0). CGA. The ordering of the set of triplets is the linear order determined by the selected order of the bases (C.2). y. v ). that is. AGU. GCU. that is. Hence. Consequently. AUA. 0. 0. z. CUA. For the three stop codons we select as representative the triplet UAA. UCU. GCA. U. which carries over every triplet XYZ to the triplet XYC. 0. Herein. AUU GCC. in fact. the RNA-less code. AAU AGC. that is. that is. UAG. 0. It is a four-dimensional subspace.GGU. GGA. z. CCU. GCG GUC. F : (x . UCC. and the SGC.UUA. is the set of triplets of the form XYC. We select the representation of every amino acid by only one triplet. UCA. G) so that the triplet at the left-most position is the one with the minimum value.UCG GAC. or endomorphism F (A9. CAU CGC. we derive encoding functions for the Extended RNA code. UAU UGC. UUG AUG CAC. F ◦ F = F . An auxiliary linear function for the different encoding functions In Table 3.1.

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

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

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

F4 . associated to the vectorial structure. as well as a combinatorial geometry. F2 . as a vector of the six-dimensional vector space (Z2 )6 . In Table 3.6. A graphical representation of the local encoding functions F1 .g. we derive an algebraic function. The general encoding function for amino acids and stop codons in the hypercube of 64 triplets As it is well known. 1}.230 Bulletin of Mathematical Biology (2007) 69: 215–243 Encoding Function F1. F3 and F4 70 F1 F2 F3 F4 20 18 19 60 21 20 21 50 17 16 15 17 Representative codon 40 12 11 11 5 10 14 13 30 9 9 8 6 11 20 6 5 7 10 4 2 3 1 4 4 0 0 1 10 20 30 40 50 60 70 Codon (lexicographic order) Fig. hence these codons have two or three representative codons. 61 out of the 64 triplets code for the 20 primary amino acids that are the building blocks of proteins. A ↔ 01. F3 and F4 . 7 Local encoding functions F1 . one representative codon but some of them appear in two (e. F2. Herein. Ser appears when F1 . It is done by means of the assignment C ↔ 00. F2 . Pro appears when F2 and F4 are applied) or even three (e. This correspondence involves the addition operation and the vectorial algebraic structure in the set of triplets. also called six-dimensional hypercube. which leads to the representation of each triplet as a sextuple of zeros and ones. We have deﬁned in the set of 64 triplets a structure of a vector space over the binary ﬁeld Z2 = {0. are applied) of the subsets. the triplets . Then a given amino acid will appear at different ordinate values. which assigns to every triplet its associated amino acid or its termination mark. U ↔ 10. 3. F3 and. that is.g. G ↔ 11. F2 and F4 .

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

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

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

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

To our knowledge. Concept of afﬁne space and its dimension Definition A. to every subset of the form v + W. ´ 1999. respectively.1.1.2. the primaeval RNY code was already frozen and that it evolved like a replicating and growing icicle. A.1. In other words. in part. is unique.Bulletin of Mathematical Biology (2007) 69: 215–243 235 acids of the RNY are also coded by triplets that appear in the SRF. or linear variety contained in V . also called a K-vector space. as well as with the concept of a linear transformation or linear endomorphism of a vector space. 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. (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. we can say that considering the symmetries of both the Extended RNA code and the RNA-less code.. Mathematical and biological background We assume that the reader is familiar with the concept of a vector space over a scalar ﬁeld K. The set v + W is also called a coset or adjoint class of the subgroup W. When two vectors u and v deﬁne the same linear variety. In the context of the frozen concept.. a single mRNA can be translated in three different reading frames which encode three trans-activators that are required for late transcription (Keck et al. for a linear variety. Given the uneven degeneracy of the genetic code it is appealing that the general encoding function is almost linear. 1990). this is the ﬁrst time in which the SGC is expressed as a mathematical function which maps each triplet onto its corresponding amino acid or stop signal. Appendix A. In the search of producing synthetic life in the laboratory (Hutchinson III et al. and it contains the element v . In the vaccinia virus. We recall that every vector space is an abelian group for the addition operation. 1990). to any of the three four-dimensional hypercubes of the Extended RNA code. and with that of vector subspace of a vector space.1. The dimension of a linear variety is deﬁned as the dimension of its associated vector subspace. it means that their . but the vector v may be any of the elements of the set v + W. genes are transcribed in a frame-shift fashion (Keck. where W is a vector subspace and v is a ﬁxed vector.. Remarks. for a ﬁxed subspace W. The phenotypic graph of amino acids is also a novel ﬁnding whose image resembles. Szathmary. et al. a single messenger RNA (mRNA) is able to encode three different proteins because messages contain three distinct putative translation initiation sites. Konecny et al. The subspace W is called the associated vector subspace of the linear variety v + W. For a vector space V we call afﬁne subspace of V . The associated vector subspace W. Definition A. and there are new triplets which altogether specify nine new amino acids.

such that u − v ∈ W. provided of a coordinate system. they are parallel. The other afﬁne subspaces. all the unitary sets are afﬁne subspaces of dimension zero. xn ). The elements xi . the identity v + W = W is obtained. . deﬁned as the set of all the ordered pairs (u. . Thus.2. The afﬁne subspaces of dimension 1 are called the lines of the geometry and those of dimension 2 are called the planes of the geometry. the vectors are represented as n-tuples (x1 . Definition A. .236 Bulletin of Mathematical Biology (2007) 69: 215–243 difference vector u − v belongs to W. .2. 1. . x2 . . Hence. that is. 0). A. a line parallel to a plane. . . . Then we say that the space has been coordinatized. . respectively. 2. . etc.. .3. The isomorphism may be deﬁned by the matching of any of the bases of V with the canonical bases of Kn . The concept of n-dimensional hypercube Let us consider a vector space of the form V = Kn .2. . . is the solution set of a linear system . The only afﬁne subspace of dimension n is the whole space V . In this case. the set or family of all the afﬁne subspaces or linear varieties contained in V . 0.2.2. or U is a subspace of W. and plane in a geometry. they form a base of the vector spaces. As it is well known. The concept of parallelism Definition A. the k-dimensional for 2 < k < n. If we take v as the null vector 0. 1) . the afﬁne subspaces of dimension 0.3. every n-dimensional K-vector space is isomorphic to the space Kn . is the ordered set of vectors e1 = (1. are the equivalent classes of the equivalence relation RW . . 0. that is. or as some vector which belongs to W. Definition A. e2 = (0. also called points of the associated geometry. 0. The linear varieties or afﬁne subspaces. all the vector subspaces are also afﬁne subspaces. where K is a ﬁeld. it is known that every afﬁne subspace v + W. We call the associated geometry of a K-vector space V . n}. 0). . From standard courses of linear algebra. . are generically called hyperplanes of the geometry. The canonical base of Kn . the afﬁne subspaces that contain the null vector. having the same dimension. It is easy to show that these vectors are linearly independent and that they generate the space. that is. . 0. if n is its dimension. lines and planes in a geometry are. en = (0. The points. Concepts of point. We say that 2 afﬁne subspaces v + W and u + U are parallel if W is a subspace of U . if the afﬁne subspaces have the same associated subspaces. In particular if W = U . parallel planes. 1.1. line. K being a ﬁeld. for i ∈ {1.. The associated geometry to a vector space. . in fact. Consequently. and 2. v ) ∈ V × V . . . A. in an n-dimensional vector space V . where the xi are elements of K. are called the coordinates of the vector. for a ﬁxed linear subspace W. . parallel cubes. According to this deﬁnition we can talk of parallel lines.

e2 and e3 of the canonical bases. The zero-dimensional hyperfaces are the vertexes of the hypercube whereas the one-dimensional hyperfaces are the edges of it.1. the solution set of the associated homogeneous system. xn ). The concept of coordinated afﬁne subspaces Definition A.4. Definition A. the vector space Kn is called n-dimensional hypercube. . The name hypercube comes from analogy with the three-dimensional case. In this case. being p a prime number. is the so-called reduction module p. A. A coordinated afﬁne subspace of the hypercube is called a hyperface of the hypercube (Z2 )6 . which is always a nonnegative integer.5. it is also called the weight of the vector.1. represent the points which are the vertexes of a cube or regular hexahedron. It is also called an orthotope because the angle of two adjacent edges is a right angle.5. The assignment to every integer.4. a coordinated afﬁne subspace is the solution set of a system whose associated matrix is diagonal. 5 of the main text. The vectorial coordinated lines are usually called coordinated axes and the vectorial coordinated planes simply coordinated planes. Definition A. i. 0). 1}. . The dimension of W is equal to n − r . and all the edges have length 1. 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. the norm is simply the number of ones that appear in the n-tuple. Definition A.Bulletin of Mathematical Biology (2007) 69: 215–243 237 of n equations with n unknowns. In Fig.. being the linear subspace. x2 . . the six-dimensional hypercube of the 64 codons is illustrated. which are remainders of the entire division by p. where r denotes the rank of the associated matrix. The concept of norm or length of a vector in the spaces (Z p )n Let us consider a vector space of the form (Z p )n . coincide with the extremes of the vectors e1 . In the particular case where K is the binary ﬁeld Z2 = {0. one of whose vertexes is the null vector 0 = (0. The two-dimensional hyperfaces are simply the faces of the hypercube. .2. 0. where (Z p ) denotes the Galois Field of p elements. its defect remainder in the entire division by p. i. A. where the eight triplets of zeros and ones. In fact. 1973). the system with the same left parts and the right members equal to zero. the 2 2 2 ordinary sum |v | = x1 + x2 + · · · + xn . We will call norm or length of a vector v = (x1 . In the case of the binary ﬁeld (Z2 ).e. of two elements. and the vertexes which are adjacent to it. the ﬁeld of non-negative integers. .. The ndimensional hypercube (Z2 )n is a regular polytope (Coxeter.3.e.1.4.

This kind of matrices. if d( f (u).6. e2 . A transformation f of the vector space V = (Z2 )n in itself. We deﬁne the distance between the vectors v and w as the norm of the difference vector v − w . e2 . it means the minimal number of edges between the two vertexes represented by the n-tuples. v ) for every pair (u. Note that the norm or length of a vector coincides with the scalar product of the vector with itself.1. . . . v ) ∈ V × V . the so-called hypercubes. this distance is the so-called Hamming distance. i... the afﬁne coordinated lines are usually called edges. . v ) of arbitrary vectors u and v . the afﬁne coordinated subspaces are called hyperfaces of the hypercube.2. . the linear isometries will also be called permutation transforms. en . . the following equality holds: |u + v | = |u| + |v | + 2 u. every linear isometry performs a permutation on the set {e1 . In general.5. We call scalar or inner product of the vectors v = (x1 . v . the Hamming distance is the number of places in which both vectors differ. we can deﬁne the concept of scalar or inner product in the following way: Definition A. . . A. x2 . .2. and the afﬁne coordinated planes are called faces of the hypercube. . . en }.7.238 Bulletin of Mathematical Biology (2007) 69: 215–243 In the more general case of a vector space (Z p )n . It can be proved that the inner product and the length or norm are related in the following way. yn ) the integer < v. A. Transformations that preserve the distance Definition A. In vector spaces of the type (Z2 )n . w >= x1 y1 + x2 y2 + · · · + xn yn . y2 . . In the case of the hypercube (Z2 )n . being the subtraction the inverse operation of the addition in the vector space (Z p )n .6. . visualizing the space as a graph. The concept of distance in the spaces (Z p )n . f (v )) = d(u. An isometric linear transformation will be called a linear isometry of the vector space. It is easy to show that a linear isometry preserves the length or norm of every element of the space.6. . 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. and geometrically. For this reason. . Actually. is called isometric if it preserves the distance between every two elements. For every pair (u. called permutation matrices.e. Definition 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 . xn ) and w = ( y1 . widely used in coding theory and in criptology. have the property that the inverse of any of them is equal to its transposed.

a linear isometry preserves the orthogonality of any pair of orthogonal vectors of the vector space. from V to V . It can be proved that a linear isometry preserves not only the distance between vectors. being R the ﬁeld of real numbers. ti jk as the translation associated to the vector ei + e j + ek. that is. the dimension is preserved. and only in this case.. . A. 2. j . n}. in fact.1. whose dimension is less than or equal to that of v + W. For translations tv and tw . Given a vector v of a vector space V . the translation associated to the vector ei + e j . According to this deﬁnition the vectors of the canonical bases. The linear isometries of the vector space (Z p )n will also be called permutation transforms or multiple rotations. one to one. as ti j . an abelian group which is a subgroup of the symmetric group S(V ) of all the bijections of the set V with itself. It is easy to prove that every afﬁne transformation carries a linear variety v + W onto another linear variety. a multiple rotation in the vector space.7. It means that the set T of all the translations of the vector space V is a group. It is clear that an afﬁne transformation tv ◦ F is bijective. if their scalar product is equal to zero. from V onto T . . . where F is a linear transformation or linear endomorphism of the vector space and tv is the associated translation of a ﬁxed vector v . Definition A. isomorphic to the additive group of V . If the afﬁne transformation is bijective. its linear component F is also bijective. being the isomorphism. As the set (Z p )n can be viewed as a subset of the Rvector space Rn . . and only if.9.2. the composition tv ◦ tw is the translation tv+w . The concept of orthogonality in a vector space (Z p )n Definition A. we call a translation associated to the vector v . .1. we will denote as ti the translation associated to the vector ei of the canonical base.9. T is. additionally. This kind of base is usually called orthonormal base. Hence. and so on. Definition A. one to each other.e. but also the scalar product of any two vectors of the vector space. a multiple rotation in (Z p )n can also be interpreted as a multiple rotation in Rn .8. We say that two vectors v and w of the space (Z p )n are orthogonal or perpendicular. and they are. i. where i . In the hypercube (Z2 )n . if.Bulletin of Mathematical Biology (2007) 69: 215–243 239 As every permutation is a composition of pairwise disjoint cycles.8. Special notation. associated to the vectors v and w . to the bijective function tv : u → u + v . The concept of translation and afﬁne transformations in a vector space Definition A. We call afﬁne transformation to every composed function of the form tv ◦ F . given by the function: v → tv . associated to the sum v + w of both vectors. of length 1. A.9.1. every linear isometry can be interpreted as a composition of local rotations. in any vector space of the form (Z p )n are orthogonal. k ∈ {1.

0. it is a transversion. it means that the vector (0. that is.10. that is. and G= 11.3. since A and G are both purines. 1. is a (k + 1)-dimensional hyperface. that is. when u is orthogonal to W and has length h greater than 1. 0. z. Definition A.1. Transitions and transversions. every mutation may be represented by means of a translation tv . A right hyperprism of height h. that every triplet XYZ is represented as a sextuple (x . if CUA is changed to CUU. It follows.9. with h greater than 1. purine to pyrimidine or pyrimidine to purine. In this particular case the mutation is simple. since it consists in the change of only one base. and it is called a transversion when the change is from one class to another. It can be proved that the union of the k-dimensional hyperfaces v + W and ei + v + W.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 . . 0. is an afﬁne subspace. 1) and this transformation may be represented by the addition of the vector e5 = (0. In our work. Then. If the translation is to a distance equal to 1. The substitution in a triplet of one or more of its three nucleotide bases within the set {C. 0. 1. According to our representation of triplets as elements of the vector space (Z2 )6 . For instance.10. t . and the nucleotides C and U are called pyrimidines. Then. is called a right hyperprism of height h. A right hyperprism is obtained from a hyperface by the adjunction of a translation of it. The nucleotides A and G are called purines. the nucleotide A is replaced by the nucleotide G. 1) is changed to the vector (0. if the triplet CUA is changed for the triplet CUG. 1. it is a transition. v ) of zeros and ones. associated to some vector v . we have assigned to every nucleotide a numerical pair in the following way: C= 00. purine to purine or pyrimidine to pyrimidine. to a distance h greater than 1. Some biological concepts and their mathematical representation A. then the purine A is changed to the pyrimidine U. u. then the set u + v + W is also a k-dimensional hyperface. The union of the hyperfaces v + W and u + v + W. We say that the mutation is simple if the substitution involves only one base of the triplet. when ei does not belong to W. but it is not a hyperface of the hypercube. U= 10. 1. Definition A. 0). for instance. On the the other hand.2. U. 0. 0. A codon mutation is called a transition when it changes a base to a base of the same class. that is. A= 01. if the triplet CUA changes to the triplet CUG. G} is called a mutation. as an element of the hypercube (Z2 )6 . Mutations Definition A.10. associated to it. This terminology is due to their chemical composition. a (k + 1)-dimensional hyperface is obtained. 0. by substitution of the base A by the base G. y. by the translation t5 . A. 0.

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

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

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