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).

U↔ 10. U. it is convenient to select the ordering in such a way that 00 is complementary of 11. For the interested reader.. whereas the pyrimidines C and U are represented by the even numbers 0 and 2. U. U). 10.3) associated to the RNA-less code can be inserted as pairwise disjoint four-dimensional afﬁne subspaces in the six-dimensional hypercube (Coxeter. 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. The article is organized as follows. and it represents the SGC code itself. in the binary numerical system.1) as derived by concepts of combinatorial geometry. Sanchez 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.b. A3.Bulletin of Mathematical Biology (2007) 69: 215–243 217 To our knowledge. 11 to the letters C. (A. This assignment of the duplets 00. 1996. Konecny et al. (A. To this end. C. which also represent the integers 0. the three four-dimensional hypercubes and the right hyperprism (A9. given that translational and transcriptional errors were probably of great importance early in the history of life. Fig.1. an Appendix which is referred to throughout this work. we discuss our ﬁndings in terms of the origin and evolution of the SGC. second and third reading frames. we consider not a strict comma-less code as proposed by Crick et al. G↔ 11. 2005). Several authors (e.. purines A and G are represented by the odd numbers 1 and 3. In previous works (Sanchez et al. Hence. A and U are complementary to each other in double stranded DNA. First.1. G). we have only eight possible selections: (C. 1995) with the SGC. and we show that this function is an integration of the different encoding functions of the above-mentioned four sets. Given that C and G. associated to the Extended RNA code. when the machinery of protein synthesis was imprecise.. we show that each reading frame can be represented as a four-dimensional hypercube (A3.1. there has not been systematic studies that relate the RNA code (Crick. 1. A. 2. Finally. A. 2004a. (1957) but rather an RNA code which can be translated in the ﬁrst. we deﬁne an encoding function for each of the three four-dimensional hypercubes. G). We also derive the general encoding function of the SGC. Here. We hypothesize that in order to allow further evolution of the RNA genetic code. we search for symmetries and patterns in both the SGC and the RNA code. 2005). Therefore. Eigen and Schuster. observing that 64 is equal not only to 43 but also to 26 . With this ordering. Next. 5). we used the following ordered assignment of the nucleotide bases: C↔ 00. 1. (C. 1977. G. and an encoding function for the RNA-less code.g. the constraint of having an intact message in only one reading frame has to be relaxed. This encoding function adopts different forms for different subsets of codons. Jimenez-Monta no ´ et al. A. suggested to organize the codon table as a six-dimensional hypercube ´ or six-dimensional vector space over the binary ﬁeld. 1968. two algebras were presented to reﬂect the relationship between codon assigment and the physicochemical aspects of the amino acids. A1. . about the concepts of algebra and geometry is provided at the end of this article. 3. may be done in 24 = 4! ways. U. G. and. C. A↔ 01.. G. and 01 is to 10. U). 1973. 01. Interestingly.

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

not only the same central nucleotide. these pairs of codons specify the same amino acid. the vector e2 + e4 . A6. but also the same ﬁrst nucleotide. For each of the eight subsets.1). j ). the 64 triplets XYZ of the SGC. 2. The remaining 16 codons can only be obtained by mutations other than by frame-shift readings. A.1) or vectorial cube.2.1). We say that the reading and translation of sequences of codons of the form RNY deﬁnes the ﬁrst reading frame (FRF) in the RNA World. For every triplet (i . Y. only one is a vector subspace (A1. Eigen and Schuster. we may consider two subclasses. Finally. RRY is obtained from YYY by the addition of the triplet AAC. each of the subclasses is partitioned into two pairs.Bulletin of Mathematical Biology (2007) 69: 215–243 219 acids. RYY is obtained from YYY by the addition of the triplet ACC. for the set RNY we obtain the two subsets RYY and RRY. and that is the subset YYY. that . contained in the six-dimensional hypercube. associated to the canonical vector e4 . and the fourth set corresponds to the RNA-less code (YYY and RRR patterns). respectively. In most cases. the start codon. For every pair (i . 1978). In this context. we denote as ti jk the composed function ti ◦ t j ◦ tk and so on (A9. j . The set of codons of the form RNY is the union of the cubes RYY (Fig. can be partitioned into four sets of 16 triplets each. and the three known stop codons. 2004. U. From the eight subsets. where X. 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. and Z belong to the set {C. Each of the ﬁrst three sets can be partitioned into two subsets by replacing the N by Y or R. every pair consisting of triplets that have. the addition of the vector e2 which involves a transversion (A10. that is. we denote as ti j the composed function ti ◦ t j . The ﬁrst set corresponds to the so-called primaeval RNA code (RNY pattern).. 1968. The eight elements of a subset correspond to a three-dimensional coordinated afﬁne subspace (A4. 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). Each cube is obtained from the other by means of the translation t4 . Now we will consider the combinatorial geometry. The fourth set is the union of two subsets: YYY and RRR. First reading frame The 16 triplets of the RNY pattern code for eight amino acids and are considered as the primaeval RNY code (Crick. In order to describe algebraically and geometrically all sets. that is. 1b) in which every amino acid corresponds to an edge of the associated cube. each of them formed by triplets having the same central nucleotide.1). They constitute coordinated lines or edges of the hypercube. that is. we denote as ti the associated translation: v → v + ei . these subclasses correspond to coordinated planes or faces of the cube. For example. which represents transversions in the ﬁrst and the second nucleotides. which contains the null vector CCC . k). G}. 1a) and RRY (Fig.2) in the ﬁrst nucleotide. we introduce the following notational convention. This is so since the Hamming distance (He et al. For every vector ei . The other seven subsets are three-dimensional afﬁne subspaces obtained from YYY by translations (A9.1) between triplets on the same edge is 1. including AUG. an ordinary cube. and we deﬁned it as the RNA-less code or complementary code of the Extended RNA code. associated to the structure of the vector space. of the canonical base.

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

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

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

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

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

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

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

We denote this hypercube as NNC (Fig. which projects the whole hypercube onto the cube RNC . z. e4 . image of the function F . representing the FRF of the primaeval RNA code. This representative triplet in the cube RNC . system u = 0.Bulletin of Mathematical Biology (2007) 69: 215–243 227 Fig. of the linear transformation F : XYZ → XYC . in fact. y. or else while preserving the ﬁrst and second nucleotides. 6). e2 . we deﬁne a function F1 . v which are the components of a generic vector in (Z2 )6 . u. the function F1 assigns to each of the 16 triplets of the set RNY the same triplet if it ends with C. U. The function F1 is the restriction. is the intersection of the four-dimensional hypercubes RNY . of the whole six-dimensional hypercube. is selected. Hypercube NNC image of the function F . Note that the blue three-dimensional cube is image of F1 . In other words. 6 The hypercube NNC . v = 0. the third one is changed to C if it is U. A. This set is a four-dimensional coordinated vector subspace. 3.2. t . according to the linear order of the set of triplets as derived from the selected ordering in the set {C. This means that for every edge associated to the same amino acid. to the hypercube RNY. e3 . which projects it onto its subspace of triplets of the form NNC. The triplet that is selected as the canonical representative of each amino acid is the one which belongs to the cube RNC . image of the function F1 in the set RNY. if compared with the other representative. G}. with unknowns x . a four-dimensional hyperface of the whole six-dimensional hypercube generated by the canonical vectors e1 . is the minor. Note that the cube RNC . 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.

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

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

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

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

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

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

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

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

deﬁned as the set of all the ordered pairs (u. . n}. or U is a subspace of W. The concept of n-dimensional hypercube Let us consider a vector space of the form V = Kn .1. Consequently. all the unitary sets are afﬁne subspaces of dimension zero. such that u − v ∈ W. that is. . 1. 0. 0). the afﬁne subspaces that contain the null vector. As it is well known. respectively. . where the xi are elements of K. If we take v as the null vector 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.. 0). Then we say that the space has been coordinatized. the afﬁne subspaces of dimension 0. The isomorphism may be deﬁned by the matching of any of the bases of V with the canonical bases of Kn . are generically called hyperplanes of the geometry. . . parallel planes. is the ordered set of vectors e1 = (1. In this case. that is. . In particular if W = U . also called points of the associated geometry. From standard courses of linear algebra. The linear varieties or afﬁne subspaces. . . 2. . lines and planes in a geometry are. etc. for i ∈ {1. in an n-dimensional vector space V . all the vector subspaces are also afﬁne subspaces.2.2.3. Definition A. Hence. Definition A. they are parallel. en = (0. they form a base of the vector spaces. having the same dimension. the identity v + W = W is obtained. is the solution set of a linear system . . The elements xi . . Concepts of point. and plane in a geometry. x2 . parallel cubes. The other afﬁne subspaces.2. . The associated geometry to a vector space. are the equivalent classes of the equivalence relation RW . Thus. 0. if n is its dimension. . . v ) ∈ V × V . A. . the k-dimensional for 2 < k < n. or as some vector which belongs to W. provided of a coordinate system. 0. every n-dimensional K-vector space is isomorphic to the space Kn . It is easy to show that these vectors are linearly independent and that they generate the space. The concept of parallelism Definition A. A. We say that 2 afﬁne subspaces v + W and u + U are parallel if W is a subspace of U . .3.. 1. According to this deﬁnition we can talk of parallel lines. e2 = (0. 1) . that is. . K being a ﬁeld. in fact.2.2. where K is a ﬁeld. . are called the coordinates of the vector. for a ﬁxed linear subspace W.236 Bulletin of Mathematical Biology (2007) 69: 215–243 difference vector u − v belongs to W. xn ). The points. 0. the vectors are represented as n-tuples (x1 . the set or family of all the afﬁne subspaces or linear varieties contained in V . The canonical base of Kn . if the afﬁne subspaces have the same associated subspaces. . and 2. The only afﬁne subspace of dimension n is the whole space V . line. We call the associated geometry of a K-vector space V . a line parallel to a plane. it is known that every afﬁne subspace v + W. . .

1. A coordinated afﬁne subspace of the hypercube is called a hyperface of the hypercube (Z2 )6 .. In fact.. of two elements. A. In the particular case where K is the binary ﬁeld Z2 = {0. e2 and e3 of the canonical bases. which is always a nonnegative integer. The concept of coordinated afﬁne subspaces Definition A. the six-dimensional hypercube of the 64 codons is illustrated.2.4.e. 5 of the main text. In this case. 1}. the vector space Kn is called n-dimensional hypercube. where the eight triplets of zeros and ones. .3. where (Z p ) denotes the Galois Field of p elements. the ﬁeld of non-negative integers. and the vertexes which are adjacent to it. a coordinated afﬁne subspace is the solution set of a system whose associated matrix is diagonal. The zero-dimensional hyperfaces are the vertexes of the hypercube whereas the one-dimensional hyperfaces are the edges of it. 0). Definition A. one of whose vertexes is the null vector 0 = (0. Definition A. We will call norm or length of a vector v = (x1 . represent the points which are the vertexes of a cube or regular hexahedron. being p a prime number. the norm is simply the number of ones that appear in the n-tuple.1. where r denotes the rank of the associated matrix. . is the so-called reduction module p. 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 . 0. The ndimensional hypercube (Z2 )n is a regular polytope (Coxeter. x2 . A. . In a vector space Kn we will call coordinated afﬁne subspace to every afﬁne subspace whose associated subspace is generated by one or some vectors of the canonical base.1. coincide with the extremes of the vectors e1 . 1973). i. In Fig. it is also called the weight of the vector.Bulletin of Mathematical Biology (2007) 69: 215–243 237 of n equations with n unknowns. The dimension of W is equal to n − r . which are remainders of the entire division by p. xn ). The vectorial coordinated lines are usually called coordinated axes and the vectorial coordinated planes simply coordinated planes. and all the edges have length 1.e. its defect remainder in the entire division by p.4.5.4. the system with the same left parts and the right members equal to zero. . The name hypercube comes from analogy with the three-dimensional case. i. Definition A. In the case of the binary ﬁeld (Z2 ). the 2 2 2 ordinary sum |v | = x1 + x2 + · · · + xn . It is also called an orthotope because the angle of two adjacent edges is a right angle.5. The assignment to every integer. being the linear subspace. . the solution set of the associated homogeneous system. The two-dimensional hyperfaces are simply the faces of the hypercube.

6. the afﬁne coordinated subspaces are called hyperfaces of the hypercube. we can deﬁne the concept of scalar or inner product in the following way: Definition A. e2 . called permutation matrices.1. For every pair (u. widely used in coding theory and in criptology. Actually. A. is called isometric if it preserves the distance between every two elements. w >= x1 y1 + x2 y2 + · · · + xn yn . . e2 . . . For this reason.. It is easy to show that a linear isometry preserves the length or norm of every element of the space.. the linear isometries will also be called permutation transforms.2.5. v ) ∈ V × V . the so-called hypercubes. it means the minimal number of edges between the two vertexes represented by the n-tuples. i. being the subtraction the inverse operation of the addition in the vector space (Z p )n . . v .7. yn ) the integer < v. Note that the norm or length of a vector coincides with the scalar product of the vector with itself. the following equality holds: |u + v | = |u| + |v | + 2 u. This kind of matrices. xn ) and w = ( y1 . en }. In the case of the hypercube (Z2 )n . In vector spaces of the type (Z2 )n . v ) for every pair (u. An isometric linear transformation will be called a linear isometry of the vector space. . 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 . . . . x2 . en .6. In general. 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 . visualizing the space as a graph.e. this distance is the so-called Hamming distance. A. We call scalar or inner product of the vectors v = (x1 .6. . and the afﬁne coordinated planes are called faces of the hypercube. and geometrically. Definition A. if d( f (u). The concept of distance in the spaces (Z p )n . . Transformations that preserve the distance Definition A. . We deﬁne the distance between the vectors v and w as the norm of the difference vector v − w . the Hamming distance is the number of places in which both vectors differ. the afﬁne coordinated lines are usually called edges. 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. It can be proved that the inner product and the length or norm are related in the following way. f (v )) = d(u. every linear isometry performs a permutation on the set {e1 . A transformation f of the vector space V = (Z2 )n in itself. y2 . . .2. . have the property that the inverse of any of them is equal to its transposed.

The linear isometries of the vector space (Z p )n will also be called permutation transforms or multiple rotations. to the bijective function tv : u → u + v . It means that the set T of all the translations of the vector space V is a group. the translation associated to the vector ei + e j . . If the afﬁne transformation is bijective. The concept of orthogonality in a vector space (Z p )n Definition A.2.8. k ∈ {1. .. of length 1. if their scalar product is equal to zero. whose dimension is less than or equal to that of v + W. but also the scalar product of any two vectors of the vector space.7. every linear isometry can be interpreted as a composition of local rotations. a multiple rotation in (Z p )n can also be interpreted as a multiple rotation in Rn .9. additionally. . The concept of translation and afﬁne transformations in a vector space Definition A. Given a vector v of a vector space V . isomorphic to the additive group of V . We say that two vectors v and w of the space (Z p )n are orthogonal or perpendicular. if. where i . and they are. and only if. the dimension is preserved.9.1. as ti j . It can be proved that a linear isometry preserves not only the distance between vectors.1. we will denote as ti the translation associated to the vector ei of the canonical base. It is easy to prove that every afﬁne transformation carries a linear variety v + W onto another linear variety. from V to V . A.8. . n}. being R the ﬁeld of real numbers. For translations tv and tw . associated to the sum v + w of both vectors. Special notation.9. a linear isometry preserves the orthogonality of any pair of orthogonal vectors of the vector space. j .1. Definition A. As the set (Z p )n can be viewed as a subset of the Rvector space Rn . According to this deﬁnition the vectors of the canonical bases. A. an abelian group which is a subgroup of the symmetric group S(V ) of all the bijections of the set V with itself. the composition tv ◦ tw is the translation tv+w . 2. in any vector space of the form (Z p )n are orthogonal. ti jk as the translation associated to the vector ei + e j + ek. its linear component F is also bijective. It is clear that an afﬁne transformation tv ◦ F is bijective.Bulletin of Mathematical Biology (2007) 69: 215–243 239 As every permutation is a composition of pairwise disjoint cycles. from V onto T . . and so on. i. associated to the vectors v and w .e. being the isomorphism. In the hypercube (Z2 )n . one to one. that is. we call a translation associated to the vector v . in fact. one to each other. T is. given by the function: v → tv . a multiple rotation in the vector space. Definition A. Hence. We call afﬁne transformation to every composed function of the form tv ◦ F . and only in this case. where F is a linear transformation or linear endomorphism of the vector space and tv is the associated translation of a ﬁxed vector v . This kind of base is usually called orthonormal base.

The union of the hyperfaces v + W and u + v + W. is an afﬁne subspace. every mutation may be represented by means of a translation tv .3. 1. 1) and this transformation may be represented by the addition of the vector e5 = (0. since it consists in the change of only one base. it is a transition. 0. A codon mutation is called a transition when it changes a base to a base of the same class. that is. associated to it. that is. is a (k + 1)-dimensional hyperface. In this particular case the mutation is simple. It can be proved that the union of the k-dimensional hyperfaces v + W and ei + v + W.10. Definition A. associated to some vector v . 0. The substitution in a triplet of one or more of its three nucleotide bases within the set {C. In our work. u. . and G= 11. z. Definition A. 0).2. v ) of zeros and ones. then the set u + v + W is also a k-dimensional hyperface. by substitution of the base A by the base G. but it is not a hyperface of the hypercube. This terminology is due to their chemical composition. If the translation is to a distance equal to 1. Then. 1. For instance. U. and the nucleotides C and U are called pyrimidines. The nucleotides A and G are called purines. it means that the vector (0. 0. A right hyperprism of height h. 0. it is a transversion. A. t . 1. 0. Some biological concepts and their mathematical representation A. and it is called a transversion when the change is from one class to another. On the the other hand. then the purine A is changed to the pyrimidine U. as an element of the hypercube (Z2 )6 . by the translation t5 . purine to pyrimidine or pyrimidine to purine. G} is called a mutation. when u is orthogonal to W and has length h greater than 1. that every triplet XYZ is represented as a sextuple (x . Mutations Definition A. We say that the mutation is simple if the substitution involves only one base of the triplet. 1. a (k + 1)-dimensional hyperface is obtained. y. the nucleotide A is replaced by the nucleotide G. since A and G are both purines. U= 10.10. 0. According to our representation of triplets as elements of the vector space (Z2 )6 . we have assigned to every nucleotide a numerical pair in the following way: C= 00. with h greater than 1.1. that is. that is. to a distance h greater than 1. when ei does not belong to 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 .10. purine to purine or pyrimidine to pyrimidine. for instance. A right hyperprism is obtained from a hyperface by the adjunction of a translation of it. 0. is called a right hyperprism of height h. It follows.9. 1) is changed to the vector (0. A= 01. Transitions and transversions. if CUA is changed to CUU. 0. Then. if the triplet CUA is changed for the triplet CUG. if the triplet CUA changes to the triplet CUG.

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

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

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