Codes of biosequences

   

   E. N. Trifonov

   Genome Diversity Center, Institute of Evolution, University of Haifa, Haifa 31905, Israel

   trifonov@research.haifa.ac.il

   

   Abstract

   

         Contrary to common belief that the nucleotide sequences only encode proteins, there are
   numerous additional codes, each code of a different nature. The codes, at DNA, RNA, and
   protein sequence levels, are superposed, that is, the same nucleotide in a given sequence may
   be simultaneously involved in several different encoded functions, at different levels. Such
   coexistence is possible due to degeneracy of the messages present in the sequence. Protein
   sequences are degenerate as well: involved not only in the functions related to the protein,
   but also adjusting to sequence requirements at the DNA level. 

   

   

   

   Key words: sequence codes; overlapping messages; chromatin code; DNA shape code; translation
   framing code; protein binary code; proteomic code; interactions between codes

   

   Introduction

   

   All manifestations of life, from elementary biomolecular interactions to human behavior, are
   tightly associated if not in full command of sequence-specific interactions. Nucleic acid or
   protein sequence patterns involved in the molecular or higher level functions stand for the
   sequence codes of the functions. The genome that carries or encodes all these sequence
   patterns is, thus, a compact, intricately organized, informational depot. To single out all
   major sequence codes and to trace them in action may be viewed as the major challenge of
   modern molecular biology, sequence biology.

   The nucleotide sequences, thus, not only encode proteins as an inexperienced reader of
   textbooks would think. Various sequence instructions are read from the DNA, RNA, or protein
   molecule each in its own way, via one or another specific molecular interaction or a whole
   network of the interactions. In the triplet code the reading device is the ribosome. In gene
   splicing the sequence signals are recognized by the spliceosome. There are also numerous
   relatively simple sequence*specific DNA*protein and RNA-protein interactions, where the
   respective sequences are read by a single protein.

   After the triplet code was spectacularly cracked (Khorana et al. 1966; Nirenberg et al. 1966;
   Ochoa et al. 1963), the impact of this event was such that nobody could even think of other
   possible codes. The triplet code was even called "genetic code", in other words the only code,
   not leaving any room for doubts. And all early history of bioinformatics revolved around this
   single code (Trifonov 2000a). Yet, already in 1968 R. Holliday noted almost en passant that,
   perhaps, recombination signals in yeast may reside on the same sequence that encodes proteins
   (Holliday 1968). This remark not only introduced the notion of other possible codes, but also
   the overlapping of different codes on the same sequence. The existence of codes, other than
   the classical translation triplet code, is already suggested by degeneracy of the triplet code
   (Schaap 1971). Freedom in the choice of codons allows significant changes in the nucleotide
   sequence without changing the encoded protein sequence. This makes it possible, in principle,
   to utilize the interchangeable bases of the mRNA sequence for some additional, different
   codes. In this case, the codes would coexist in interspersed form as mosaics of two or more
   "colors". It is known today that a more general and wide-spread case is when the codes
   literally overlap so that some letters in specific positions of a given sequence (nucleotides
   or amino acids) are simultaneously involved in two or more different codes (sequence
   patterns). Such is the case with the coexisting triplet code and chromatin code - sequence
   instructions for nucleosome positioning (Trifonov 1980; Mengeritsky and Trifonov 1983). This
   was the first demonstration of the actual existence (Trifonov 1981) of the hypothetical
   overlapping codes. Sequences that do not encode proteins, despite their traditional
   classification as non*coding, carry some important messages (codes) as well. Especially
   striking are the cases of sequence conservation in the non-coding regions (Koop and Hood
   1994), suggesting that the so-called non-coding sequences are associated with some function.

   Amongst known general sequence codes, other than the triplet code, are transcription signals
   (transcription code) in promoters such as TATAAA*box in eukaryotes, and TATAAT* and
   TTGACA*boxes in bacteria coding for initiation of transcription. Another broadly known
   sequence code is the gene splicing code, the GT*AG rule (Breathnach and Chambon 1981) and some
   sequence preferences around the intron*exon junctions. A complex set of sequence rules
   describes details of DNA shape important for DNA*protein interactions and DNA folding in the
   cell.

   At the level of amino acid sequences the most important is the protein folding code, which is
   not yet described as sequence pattern. One can single out the modular component of the folding
   code - organization of the globular proteins as linear succession of the modules in the form
   of loops of 25 - 30 residues closed at the ends by interactions between hydrophobic residues
   (Berezovsky et al. 2000; Berezovsky and Trifonov 2002). The 3-D structure of proteins appears
   to be encoded largely by a binary code (Trifonov et al. 2001; Trifonov 2006; Gabdank et al.
   2006) that, essentially, reduces the 20-letter alphabet to only two letters, for non-polar and
   polar residues (more accurately - residues encoded by codons with pyrimidine or purine in the
   middle). The binary code also suggests the ancestral form for any given sequence.

   As the carriers of instructions the biological sequences may be considered a language. Indeed,
   according to an appealing definition of Russian philosopher V. Nalimov (1981), language is a
   communication  tool to carry instructions to the operator at the receiving end. Such languages
   as computer programs (frequently called "codes" as well) and written (spoken) human languages
   convey instructions expressed in the form of one code, for one reading device that takes
   consecutively letter by letter, word by word, until the transmitted command is fully uttered.
   As mentioned above, a unique property of the biological sequences is the superposition of the
   codes they carry. That is, the same sequence is meant to be read by several reading devices,
   each geared to its own specific code. Many cases of such overlapping are known (Trifonov 1981;
   Normark et al. 1983). The overlapping is possible due to degeneracy of the codes. There is, of
   course, an informational limit for such superposition, when the freedoms of degeneracy become
   insufficient to accommodate additional messages without loss of quality of many or all other
   messages present.

   

   Hierarchy of the codes

   

   The commonly considered information flow from DNA to RNA and to protein is accompanied by
   massive loss of the sequences involved. Indeed, not all DNA is transcribed, nor is the whole
   mass of RNA transcripts translated. This is especially obvious in eukaryotic genomes that
   contain large intergenic regions, and large intervening sequences that are passed from DNA to
   pre-mRNA. Is that loss of sequences also a loss of information? The multiplicity of the codes
   and their superposition suggest that some information is lost, indeed, together with those
   sequences that are not transcribed and not translated. In other words, DNA carries the
   sequence codes, serving at the DNA level, of which some are transferred to pre-mRNA. The
   sequences of the transcripts carry codes serving at RNA level, of which some are passed to the
   protein sequences, via mRNA. One, thus, has to consider the codes characteristic for the three
   sequence levels, hierarchically.

   One could think of yet higher level codes, beyond the purely molecular level. Among them would
   be organ/tissue-specific codes, i. e. genomic sequence features characteristic for one or
   another physiological function. These could be specifically placed tandem repeats, dispersed
   repeats, amplified genes or whole groups of genes. One could also imagine "personal code(s)" -
   various sequence details responsible for individual traits, such as distinct facial features
   (Fondon and Garner 2004) and mimic (Peleg et al. 2006), body set, favorite postures and
   gestures and, perhaps, personal behavioral traits. Well documented existence of
   population-specific genetic diseases and disorders indicates that there are also sequence
   features responsible for ethnicity traits. These may include also specific sequence
   polymorphisms and, perhaps, some "guest" sequences present in one ethnical group and absent in
   others. The higher level codes are likely to become a major focus of molecular medicine in
   coming decades. In the mean time the sequence codes of molecular levels are still struggling
   to make it from singular to plural.

   

   DNA level codes

   

   The DNA structure is not monotonously uniform. It is modulated by the sequence*dependent local
   deviations from standard geometry, which may accumulate, for example, to a net DNA curvature
   (Trifonov and Sussman 1980). Geometry of every base*pair step in the simple wedge model is
   described by three angles * wedge roll, wedge tilt, and twist. By following the sequence and
   deflecting the DNA axis at every step, according to the wedge and twist angles from the table
   of the dinucleotide codons (Bolshoy et al. 1991; Trifonov 1991), one can calculate the
   predicted path of DNA axis - its local shape for any given sequence (Shpigelman et al. 1993).
   Hence, DNA shape code.

    The chromatin code is a set of rules directing sequence*specific positioning of the
   nucleosomes. Sequence*dependent deformational anisotropy (bendability) of DNA appears to be an
   underlying principle of the nucleosome sequence specificity (Trifonov 1980). As the strands of
   the nucleosome DNA follow the path of the deformed DNA duplex, they pass through inner contact
   points with histones (interface positions) and outward points (exposed to nucleoplasm).
   Various sequence elements, that prefer the inner or outward positions would, thus, ideally,
   reappear in the sequence at the distances that are multiples of nucleosome DNA period. Indeed,
   the sequence periodicity is the most conspicuous feature of the nucleosome DNA sequences
   (Trifonov and Sussman 1980). According to the latest updates (Cohanim et al. 2005, 2006a;
   Kogan and Trifonov 2006; Trifonov et al. 2006a), there are at least three major periodical
   patterns in the nucleosome DNA: counter-phase AA/TT pattern, counter-phase GG/CC pattern (both
   combined in RR/YY pattern), and in-phase AA/TT pattern. Several other possible patterns are
   discussed in literature (reviewed in Kiyama and Trifonov 2002; Segel et al. 2006).

   An important issue in the elucidation of the chromatin sequence code is mandatory weakness of
   the nucleosome positioning sequence signal. This is required by the necessity of unfolding the
   nucleosomes during template processes. That is, the DNA complexes with the histone cores in
   the nucleosomes should be of marginal stability only. Accordingly, the sequence elements
   associated with the DNA bendability should be rather scarce in the nucleosome DNA sequence,
   especially those elements that are strong contributors to the bendability. Regrettably, it
   makes the deciphering of the nucleosome positioning code quite a challenge.

   One of the factors influencing the nucleosome positioning is sterical exclusion of the
   nucleosomes by other nucleosomes, neighbors in 3-D space (Ulanovsky and Trifonov 1986). The
   most obvious sterical rule is the rule of linkers, first formulated and experimentally
   observed by Noll et al. (1981). Since every extra base pair in the linker causes rotation of
   the nucleosome around the axis of the linker by ~34°, the rotation may result in a sterical
   clash between the nucleosomes connected by the common linker. This effect, indeed, is observed
   at short linkers. It is expressed in preferential appearance of the linkers of lengths about 5
   to 11, 16 to 21, and 26 to 31 bases (Noll et al. 1981; Mengeritsky and Trifonov 1983;
   Ulanovsky and Trifonov 1986; Cohanim et al. 2006a). Intermediate linker lengths are forbidden
   due to the sterical clashes ("interpenetration" of the nucleosomes). The rule of linkers,
   thus, is an important part of the chromatin code.

   

   RNA level codes

   

   Those messages contained in the transcribed DNA are passed to RNA. The transcribed DNA, thus,
   contains overlapping messages of both DNA and RNA levels. The major mRNA level message is the
   classical triplet code * RNA-to-protein translation code. The chapters about this code appear
   in every textbook on molecular biology, and it will not be described here. 

   Eukaryotic transcripts also carry the RNA splicing code. This code is only poorly described
   (Breathnach and Chambon 1981; Mount 1982), so that existing sequence-based algorithms are not
   sufficient for detection of the splice sites in the sequences with as high precision as in
   natural splicing process.

         Overlapping with the protein-coding message, sequence of codons-triplets, is the
   universal 3-base periodicity with the consensus (G-nonG-N)[n]  (Trifonov 1987) or, more
   accurately, (GCU)[n]  (Lagunez-Otero and Trifonov 1992). Since the mRNA binding sites in the
   ribosome possess a complementary periodicity (xxC)[n], with obligatory cytosines complementary
   to the frequent guanines of the first codon positions in mRNA, these 3-base periodicities have
   been interpreted as a device to maintain correct reading frame during translation of mRNA *
   the framing code (Trifonov 1987). As described below, the periodical pattern (GCU)[n] in mRNA
   appears to be a fossil of very ancient organization of codons (Trifonov and Bettecken 1997).

         The usage of codons corresponding to the same amino acid is known to be different for
   different organisms and even different genes. Among the alternative codons the rare codons are
   of special interest. Their occurrence along the mRNA sequence is not random. It is shown, for
   example, that clusters of infrequently used codons in prokaryotic mRNA often follow at a
   distance about 150 triplets from one another. This is interpreted as translation pausing code,
   to slow down the translation after a protein domain (fold) is synthesized: to give the newly
   synthesized chain sufficient time for its proper folding (Makhoul and Trifonov 2002).

   

   Codes of protein sequences

   

   According to common belief, the protein sequence carries instructions on how the polypeptide
   chain folds, for the reliable performance of respective function of the protein, encoded in
   the sequence as well. At the same time, it is well known that proteins with the same fold and
   the same function may have rather different sequences. As in the case of the triplet code,
   this degeneracy of the protein sequence may allow incorporation in the same sequence of some
   additional messages.

   The protein folding code is a major challenge for the protein structure community. There are
   plenty of sophisticated approaches offering partial solutions of the problem, but the
   conclusive sequence rules for protein folding are still to be found.

   An apparent major obstacle is estimated colossal time required for the unfolded polypeptide
   chain to go through all intermediate states until the final native fold structure is reached -
   the so-called Levinthal paradox. Some trick of nature, a special sequence organization should
   be there, in the protein sequences, to ensure the folding in realistic time of milliseconds to
   seconds. One possible way out is suggested by modular organization of the protein folds
   (Berezovsky and Trifonov 2002). Indeed, if the chain length of the module is 20-30 amino acid
   residues, the time required for its folding fits well to the realistic limits. And, as
   numerous recent studies demonstrate, globular proteins are built of such modules of standard
   size 25-30 residues in form of closed loops (Berezovsky et al. 2000; Trifonov and Berezovsky
   2003; Berezovsky et al. 2003a, 2003b; Aharonovsky and Trifonov 2005; Sobolevsky and Trifonov
   2006). 

         The modular structure of proteins suggests a principally new, compressed way of
   presentation of amino acid sequences as rather  sequences of the modules, descendants of the
   early sequence/structure/function prototypes (Berezovsky et al. 2003a, 2003b), in a new
   alphabet of the prototypes. This would represent the proteomic code contained in the amino
   acid sequences. The prototype modules, then, would appear as the codons of the proteomic code.

   

   Fast adaptation code

   

   This code resides and functions in all three types of genetic sequences. It is believed to be
   responsible for special type of quick, significant changes in the sequences, apparently, in
   response to environmental changes. It involves the most variable sequences * simple tandem
   repeats of the structure (AB*MN)[n]. Remarkably, the information carried in the sequences
   resides not as much in the sequence AB*MN of the repeating unit, as rather in the copy number
   n of the repeats (Trifonov 1989, 2004). Indeed, after the spontaneous change in the repeating
   sequence, its extension or shortening, the sequence in brackets stays intact while the copy
   number n becomes larger or smaller, respectively. Since the repeats are involved in gene
   expression in one or another way, the change of n results in the modulation of gene
   activities, as a response to environmental challenges, and thus in fast adaptation (Trifonov
   1989, 1990, 1999, 2004; Holliday 1991; King 1994; Künzler et al. 1995, King et al. 1997). An
   important faculty of this mechanism is an apparent directionality of the mutational changes of
   this type (Trifonov 2004). Indeed, small variations in the n values corresponding to repeats
   serving genes irrelevant to a given environmental stress do not change the expression patterns
   of these genes. On the contrary, if relevant, responsive genes are involved, the copy numbers
   of the respective repeats become subject of systematic selection towards better repeat copy
   number (better gene expression) patterns. The relevant genes (but only relevant ones) become,
   thus, retuned (King et al. 1997; Trifonov 1999).

   

   The codes of evolutionary past

   

   Every sequence has its evolutionary history, and those sequences  or sequence fragments, that
   have been successful in the earliest times of molecular evolution, are, perhaps, still around
   in hidden form or even unchanged since those times. The proteomic code described above is an
   example of such code of evolutionary record. The modern sequence modules are not the same as
   their ancestral prototypes, but a certain degree of resemblance to the ancestors is conserved
   allowing classification of present-day modules.

   The earliest traced sequence elements go back to the very first codons, which are described as
   the triplets GGU, GCC, and their point mutational versions (Trifonov and Bettecken 1997). More
   detailed reconstruction confirmed this conclusion (Trifonov 2000b, 2004). According to the
   reconstruction of the earliest stages of molecular evolution, the very first "genes" had a
   duplex structure with complementary sequences (GGC)[n] and (GCC)[n], encoding, respectively,
   Gly[n] and Ala[n]. Thus, the mRNA consensus (GCU)[n] and the consensus (xxC)[n] of the mRNA
   binding sites in the ribosome are both fossils of the earliest mRNA sequences (Trifonov 1987;
   Lagunez-Otero and Trifonov 1992; Trifonov and Bettecken 1997).

   The size of the earliest minigenes, as it turns out, can be estimated by distance analysis of
   modern mRNA sequences (Trifonov et al. 2001). For this purpose the sequences were first
   rewritten in binary form, in an alphabet of two letters, G and A , for Gly series of amino
   acids and codons and Ala series (see above). Respective codons contain in their middle
   positions either purines (in G) or pyrimidines (in A). From the reconstructed chart of
   evolution of the codons (Trifonov 2000b, 2004), it follows that all codons of G-series are
   descendants of the GGC codon, with purine in the middle, while codons of A-series originate
   from GCC codon, with pyrimidine in the middle. If the products of very first genes had the
   structures either G[n] or A[n], of a certain size n, then after fusion of the minigenes the
   alternating patterns G[n]A[n]G[n]A[n]... may have been formed. Later mutations could, of
   course, have completely destroyed this pattern, but they did not. Analysis of large ensembles
   of the mRNA sequences showed that the pattern did survive, though in rather hidden form
   (Berezovsky and Trifonov 2001; Trifonov et al. 2001) so that the estimation of the very first
   gene size became possible - 6-7 codons encoding hexa- and hepta-peptides. This estimate is
   strongly supported by independent calculation of the sizes of the most ancient mRNA hairpins
   that arrived at the same minigene size (Gabdank et al. 2006; Trifonov et al. 2006b). Moreover,
   most conserved oligopeptide sequences, present in every prokaryotic proteome, also have the
   size of 6-9 amino acids (Sobolevsky and Trifonov 2005, Sobolevsky et al., manuscript in
   preparation).

   The ancient conservation of the middle purines and pyrimidines in the codons during the
   evolution of the codon table, actually, has very much survived till now. This is confirmed by
   an analysis of amino acid substitutions in modern proteins (Trifonov 2006; Gabdank et al.
   2006). Every modern protein sequence, thus, can be written in the A and G alphabet. Such
   presentations of modern sequences in the binary code would suggest the most ancient version of
   the sequences.

   The binary code, the mosaic of A- and G-minigenes, and the proteomic code describe various
   stages of protein evolution, from simple to more complex. Today one can also detect the next
   stage - combining the closed loop modules in the protein folds, domains.

   First, the next level is seen already in protein sizes, which appear to be multiples of
   120-150 amino acid units (Berman et al. 1994; Kolker et al., 2002). This size is a good match
   to the optimal DNA ring closure size, about 400 base pairs (Shore et al., 1981). This
   attractive numerology may well reflect original formation of modern genes and genomes by
   fusion of individual DNA circles (genome units) of this standard size (Trifonov, 1995, 2002).
   This would constitute the genome segmentation code. How this code is expressed in the sequence
   form is not yet specified, except for preferential appearance of methionines (former
   translation starts) at genome unit size distances (Kolker and Trifonov 1995).

   

   Superposition of the codes and interactions between them

   

   As most of the codes described above are degenerate, allowing alternative or sometimes even
   wrong letters here and there, they may coexist as a superposition of several codes, on the
   same sequence (reviewed in Normark et al. 1983; Trifonov 1981, 1989, 1996, 1997). The most
   spectacular case is the overlapping of the chromatin code (nucleosome positioning) with
   protein coding and gene splicing. Indeed, the alternating AA/TT nucleosome pattern is
   demonstrated to be located largely, if not fully, on those sections of the protein-coding
   regions that correspond to amphipathic α-helices (Cohanim et al. 2006a, b). The third
   positions of the codons within the region occupied by the nucleosome are responsible as well
   for the creation of the periodical AA/TT pattern. Moreover, even the encoded amino acid
   sequence is also biased to a certain degree to contribute to the nucleosome sequence pattern
   (Cohanim et al. 2006b). In addition, the nucleosomes are preferentially centered at the splice
   junctions, apparently for their protection (Denisov et al. 1997; Kogan and Trifonov 2005).
   Since the coding sequences also carry at least one more message - translation framing, the
   nucleosome sequences display superposition of at least four different codes, on the same
   sequence.

   The adjustment of the protein sequence, to contribute to the DNA sequence periodicity, both in
   prokaryotes and in eukaryotes (Cohanim et al. 2006b), is an interesting case. Apparently, on
   one hand, the 10-11 base DNA sequence periodicity is of no less importance for the cell than
   the proteins encoded in the DNA sequence. On the other hand, this example of interactions
   between the codes shows that the DNA sequence level message is projected all the way through
   mRNA to the protein sequence level. The latter one, thus, carries (reflects) the sequence
   patterns of the whole hierarchy - of DNA, RNA, and protein levels.

   A neat example of the overlapping at the level of protein sequences is the "moonlighting" of
   intrinsically unfolded proteins, IUP  (Tompa et al. 2005). That is, the same molecule of the
   IUP, the same sequence, can be involved in more than one function, thus, carrying different
   superimposed messages. Structural and functional promiscuity of the intrinsically unfolded
   proteins is carried through, perhaps, since the earliest times of molecular evolution. Highly
   structured functionally specialized proteins were not yet around, and the multi-functionality
   of simpler IUP molecules was of an obvious advantage for survival.  

   

   Is that all?

   

   There are still many non-deciphered codes around. Nature would utilize every useful
   combination of letters. This is because of eternal molecular opportunism (Doolittle 1988) that
   drives the molecules of life towards better and more diverse performance in the challenging
   conditions of the changing environment. In this struggle for survival (natural selection) and
   for better well being (opportunism) living matter developed intricate levels of complexity,
   including sequence complexity. It would be naive to say that all the codes are already known,
   as it was, indeed, naive to content oneself with the single "genetic code" 30 years ago.

   On the one hand, there are sequence biases and patterns that are still not fully explained,
   such as species-specific G+C content of genomes - genomic code (D'Onofrio and Bernardi 1992),
   and general avoidance of the CG dinucleotides. On the other hand, many of the known molecular
   functions still do not have explicit sequence descriptions, such as RNA interference (Fire et
   al. 1998) or RNA editing (Gott and Emeson 2000). The so-called non-coding sequences have the
   provocative property of being rather dispensable, though they do carry some of the codes
   described in the review (chromatin code, fast adaptation code). The famous case of the
   Fugu-fish genome, with the reduced amount of non-coding sequences in it (Aparicio et al.
   2002), is often taken as an example of a seemingly insignificant role the non-coding sequences
   play. Yet, it is known that the non-coding sequences harbor various repeats, of dispersed type
   (transposons), and tandem repeats. It is also known that transposable elements play an
   important role in evolution and adaptation (Reanney 1976). The tandem repeats serve as tuners
   of gene expression (Trifonov 1989, 2004; King et al. 1997; Fondon and Garner 2004)(see Fast
   adaptation code, above). Could it be that the Fugu-fish is in an evolutionary steady state,
   with virtually no need for adaptive sequence changes? That could be only if there are no
   environmental challenges for this species. Indeed, the small-genome Fugu-fish has a narrow
   habitat (Hinegardner 1976), living only in coral reefs with well defined fauna, around the
   islands of Japan. Thus, even dispensable sequences deserve respect, as they seem to code for
   the vital ability for adaptation.

   The conspicuously primitive simple tandem repeats are the best advocates in favor of all
   sequences, no matter how nonsensical, primitive, or even dispensable they appear. In a recent
   study (Bacolla et al. 2006), the pure purine or pyrimidine repeats are shown to be the only
   difference between human and chimpanzee sequences (over 800 large segments studied). The
   repeats are also the same, but the copy numbers of the repeat units (total lengths of the
   repeat regions) are different in these two species. Referring to the fast adaptation code
   (above), one would think that humans and chimpanzees are nearly the same species, only well
   adapted to completely different living conditions. So much for even the primitive sequences.

   The answer to the question in the title of this section, thus, is firm "No".

   

   

   

   

   

   

   

   

   References:

   

   Aharonovsky E, Trifonov EN (2005) Protein sequence modules. J Biomol Str Dyn 23:237-242

   Bacolla A, Collins JR, Gold B et al (2006) Long homopurine*homopyrimidine sequences are
   characteristic of genes expressed in brain and the pseudoautosomal region. Nucl Acids Res
   34:2663-2675 

   Berezovsky IN, Trifonov EN (2001) Evolutionary aspects of protein structure and folding.
   Molecular Biology 35:233-239

   Berezovsky IN, Trifonov EN (2002) Loop fold structure of proteins: Resolution of Levinthal*s
   paradox. J Biomolec Str Dyn 20:5-6

   Berezovsky IN, Grosberg AY, Trifonov EN (2000) Closed loops of nearly standard size: common
   basic element of protein structure. FEBS Letters 466:283-286

   Berezovsky IN, Kirzhner VM, Kirzhner A et al (2003a) Protein sequences yield a proteomic code.
   J Biomol Struct Dyn 21:317-325

   Berezovsky IN, Kirzhner A, Kirzhner VM, Trifonov EN (2003b) Spelling protein structure. J
   Biomol Struct Dyn 21:327-339

   Berman AL, Kolker E, Trifonov EN (1994) Underlying order in protein sequence organization.
   Proc Natl Acad Sci USA 91:4044-4047

   Bolshoy A, McNamara P, Harrington RE, Trifonov EN (1991) Curved DNA without AA: experimental
   estimation of all 16 wedge angles. Proc Natl Acad Sci USA 88:2312-2316

   Breathnach R, Chambon P (1981) Organization and expression of eukaryotic split genes coding
   for proteins. Ann Rev Bioch 50:349-383

   Cohanim AB, Kashi Y, Trifonov EN (2005) Yeast nucleosome DNA pattern: deconvolution from
   genome sequences of S. cerevisiae. J Biomol Str Dyn 22:687-694

   Cohanim AB, Kashi Y, Trifonov EN (2006a) Three sequence rules for chromatin. J Biomol Struct
   Dyn 23:559-566

   Cohanim AB, Trifonov EN, Kashi Y (2006b) Specific selection pressure on the third codon
   positions: contribution to 10 - 11 base periodicity in prokaryotic genomes. J Molec Evol (in
   press)

   Denisov DA, Shpigelman ES, Trifonov EN (1997) Protective nucleosome centering at splice sites
   as suggested by sequence-directed mapping of the nucleosomes. Gene 205:145-149

   D*Onofrio G, Bernardi G (1992) A universal compositional correlation among codon positions.
   Gene 110:81-88

   Doolittle RF (1988) More molecular opportunism. Nature 336:18

   Fire A, Xu S, Montgomery MK et al (1998) Potent and specific genetic interference by
   double-stranded RNA in Caenorhabditis elegans. Nature 391:806-811.

   Fondon JW, Garner HR (2004) Molecular origin of rapid and continuous morphological evolution.
   Proc Natl Acad Sci USA 101:18058-18063

   Gabdank I, Barash D, Trifonov EN (2006) Tracing ancient mRNA hairpins. J Biomol Str Dyn
   24:163-170

   Gott JM, Emeson RB (2000) Functions and mechanisms of RNA editing. Ann Rev Genet 34:499-531

   Hinegardner R (1976) Evolution of genome size. In: Ayala FJ (ed) Molecular evolution. Sinauer
   Assoc Inc, Sunderland

   Holliday R (1968) Genetic recombination in fungi. In: Peacock WJ, Brock RD (eds) Replication
   and recombination of genetic material. Australian Academy of Science, Canberra 

   Holliday R (1991) Quantitative genetic variation and developmental clocks. J Theor Biol
   151:351-358

   Khorana HG, Büchi H, Ghosh H et al (1966) Polynucleotide synthesis and the genetic code. Cold
   Spring Harb Symp Quant Biol 31:39-49

   King DG (1994) Triple repeat DNA as a highly mutable regulatory mechanism. Science 263:595-596

   King DG, Soller M, Kashi Y (1997) Evolutionary tuning knobs. Endeavor 21:36-40

   Kiyama R, Trifonov EN (2002) What positiones nucleosomes? - A model. FEBS Let 523:7-11

   Kogan S, Trifonov EN (2005) Gene splice sites correlate with nucleosome positions. Gene
   352:57-62

   Kogan SB, Kato M, Kiyama R, Trifonov EN (2006) Sequence structure of human nucleosome DNA. J
   Biomol Struct Dyn 24:43-48

   Kolker E, Trifonov EN (1995) Periodic recurrence of methionines: Fossil of gene fusion? Proc
   Natl Acad Sci USA 92:557-560

   Kolker E, Tjaden BC, Hubley R et al (2002) Spectral analysis of distributions: finding
   periodic components in eukaryotic enzyme length data. OMICS: J Integr Biol 6:123-130

   Koop BF, Hood L (1994) Striking sequence similarity over almost 100 kilobases of human and
   mouse T-cell receptor DNA. Nature Genet 7:48-53

   Künzler P, Matsuo K, Schaffner W (1995) Pathological, physiological, and evolutionary aspects
   of short unstable DNA repeats in the human genome. Biol Chem Hoppe-Seyler 376:201-211

   Lagunez-Otero J, Trifonov EN (1992) mRNA periodical infrastructure complementary to the
   proof-reading site in the ribosome. J Biomol Struct Dyn 10:455-464

   Makhoul CH, Trifonov EN (2002) Distribution of rare triplets along mRNA and their relation to
   protein folding. J Biomol Struct Dyn 20:413-420

   Mengeritsky G, Trifonov EN (1983) Nucleotide sequence-directed mapping of the nucleosomes.
   Nucl Acids Res 11:3833-3851

   Mount SM (1982) A catalogue of splice junction sequences. Nucl Acids Res 10:459-472

   Nalimov VV (1981) In the labyrinths of language: A mathematician's journey. ISI Press,
   Philadelphia

   Nirenberg M, Caskey T, Marshall R et al (1966) The RNA code and protein synthesis. Cold Spring
   Harb Symp Quant Biol 31:11-24

   Noll M, Zimmer S, Engel A, Dubochet J (1980) Self-assembly of single and closely spaced
   nucleosome core particles. Nucl Acids Res 8:21-42  

   Normark S, Bergstrom S, Edlund T et al (1983) Overlapping genes. Ann Rev Genet 17:499-525

   Ochoa S (1963) Synthetic polynucleotides and the amino acid code. Cold Spring Harb Symp Quant
   Biol 28:559-567

   Peleg G, Katzir G, Peleg O et al (2006) Hereditary family signature of facial expression. Proc
   Natl Acad Sci USA 103:15921-15926

   Reanney DC (1976) Extrachromosomal elements as possible agents of adaption and development.
   Bact Rev 40:552-590

   Schaap T (1971) Dual information in DNA and the evolution of the genetic code. J Theor Biol
   32:293-298

   Shore D, Langowski J, Baldwin RL (1981) DNA flexibility studied by covalent closure of short
   fragments into circles. Proc Natl Acad Sci USA 78:4833-4838

   Shpigelman ES, Trifonov EN, Bolshoy A (1993) CURVATURE: software for the analysis of curved
   DNA. CABIOS 9:435-440

   Sobolevsky Y, Trifonov EN (2005) Conserved sequences of prokaryotic proteomes and their
   compositional age. J Mol Evol 61:591-596

   Sobolevsky Y, Trifonov EN (2006) Protein modules conserved since LUCA. J Mol Evol 63:622-634

   Tompa P, Scasz C, Buday L (2005) Structural disorder throws new light on moonlighting. Trends
   Bioch Sc 30:484-489

   Trifonov EN (1980) Sequence-dependent deformational anisotropy of chromatin DNA. Nucl Acids
   Res 8:4041-4053

   Trifonov EN (1981) Structure of DNA in chromatin. In: Schweiger H (ed) International cell
   biology 1980-1981. Springer-Verlag, Berlin

   Trifonov EN (1987) Translation framing code and frame-monitoring mechanism as suggested by the
   analysis of mRNA and 16S rRNA nucleotide sequences. J Molec Biol 194:643-652

   Trifonov EN (1989) The multiple codes of nucleotide sequences. Bull Math Biol 51:417-432

   Trifonov EN (1991) DNA in profile. Trends Biochem Sc 16:467-470

   Trifonov EN (1990) Making sense of the human genome. In: Sarma RH, Sarma MH (eds) Structure
   and methods, vol. 1, Human genome initiative and DNA recombination. Adenine Press, New York

   Trifonov EN (1995) Segmented structure of protein sequences and early evolution of genome by
   combinatorial fusion of DNA elements. J Mol Evol 40:337-342

   Trifonov EN (1996) Interfering contexts of regulatory sequence elements. CABIOS 12:423-429

   Trifonov EN (1997) Genetic sequences as product of compression by inclusive superposition of
   many codes. Molecular Biology 31:759-767

   Trifonov EN (1999) Elucidating sequence codes: three codes for evolution. Annals NY Acad Sci
   870:330-338

   Trifonov EN (2000a) Earliest pages of bioinformatics. Bioinformatics 16:5-9

   Trifonov EN (2000b) Consensus temporal order of amino acids and evolution of the triplet code.
   Gene 261:139-151

   Trifonov EN (2002) Segmented genome: elementary units of genome  structure. Russian J Genet.
   38:659-663

   Trifonov EN (2004) The triplet code from first principles. J Biomol Struct Dyn 22:1-11

   Trifonov EN (2006) Theory of early molecular evolution: predictions and confirmations. In:
   Eisenhaber F (ed) Discovering biomolecular mechanisms with computational biology. Landes
   Bioscience, Georgetown

   Trifonov EN, Berezovsky IN (2003) Evolutionary aspects of protein structure and folding, Curr
   Opinion Struct Biol 13:110-114

   Trifonov EN, Bettecken T (1997) Sequence fossils, triplet expansion, and reconstruction of
   earliest codons. Gene 205:1-6

   Trifonov EN, Sussman JL (1980) The pitch of chromatin DNA is reflected in its nucleotide
   sequence. Proc Natl Acad Sci USA 77:3816-3820

   Trifonov EN, Kirzhner A, Kirzhner VM, Berezovsky IN (2001) Distinct stages of protein
   evolution as suggested by protein sequence analysis. J Mol Evol 53:394-401

   Trifonov EN, Kogan S, Cohanim AB (2006a) Latest on the nucleosome positioning sequence
   patterns. In: Kiyama R, Shimizu M (eds) Gene expression and chromatin. Research Signpost,
   Trivandrum (India) (in press)

   Trifonov EN, Gabdank I, Barash D, Sobolevsky Y (2006b) Primordia vita. Deconvolution from
   modern sequences. Origin Life Evol Biosph (in press)

   Ulanovsky LE, Trifonov EN (1986) A different view point on the chromatin higher order
   structure: steric exclusion effects. In: Sarma RH, Sarma MH (eds)  Biomolecular stereodynamics
   III. Adenine Press