CONTRIBUTIONS TO RNA STRUCTURE
( included in the web site http://www.lps.ens.fr/~ ninio )
TOPICS DISCUSSED HERE:
- the early days in Luzzati's laboratory - biographical notes
- molecular model for transfer RNA (1969)
- small-angle X-ray scattering work (1972)
- base-pairing statistics (1971, 1979)
SEE IN OTHER SECTIONS
- codon-anticodon recognition (accuracy of molecular processes section)
- model for the evolution of tRNA structure (origin of life section)
- topology of RNA structures (bio-mathematics section)
- prediction of RNA secondary structure (bio-informatics section)
THE EARLY DAYS - BIOGRAPHICAL NOTES
My initial education was in mathematics and engineering at Ecole Polytechnique (Paris, 1961-1963), I was seeing myself as a future engineer who would design and construct dams or bridges. Towards the end of the studies, my friend Maurice Hofnung learnt about the new openings in biological sciences. He invited me to a joint appointment with Jacques Monod, to which we went already half convinced . Monod told us about the disappointments and the rewards of scientific research. I had personal reasons at that time to engage into new activities, and Hofnung and I, under the tutorship of Monod made our first steps in biology. We spent one year studying biochemistry and genetics at the University of Paris. At the end of this year Monod advised me to join Vittorio Luzzati's laboratory at the CNRS in Gif-sur-Yvette.
Luzzati had developped a methof for studying macromolecules in solution, exploiting the information in the scattering of an X-ray beam by the solution. The theory relating the scattered intensity to the structure of the solute had been worked out by others, under the standard condition of an incident X-ray beam having a very small cross-section. Luzzati's contribution had been to extend the theory to X-ray beams with a linear cross-section (from which a much stronger signal could be recovered), and design an apparatus for collecting x-ray scattering data from linearly collimated X-ray beams. He proposed to me to work in his laboratory using this technique, and applying it to various molecules of interest. One assumption, behind such a structural work was that the folding of a macromolecule in solution could be significantly different from the folding derived from crystallographic studies, and another strong assumption was that the studies of the macromolecules in solution would reveal a multiplicity of conformational changes that would be associated with various stages of the biological functionning of the molecule.
I accepted at once without further inquiries. I expected that I could make a judicious use of my mathematical skills in the X-ray work and was excited at the prospect of working on transfer RNA, the still mysterious key molecule of the genetic code. I took with me, on the 1964 summer vacations Guinier's treatise of crystallography (over 600 pages) and conscientiously read it on the beaches. I was totatly ignorant of the publication system, and of what a carreer in scientific research implied. I had read about Evariste Galois who worked at home, and Albert Einstein who worked in a patent's office. So, I had a job in a lab and was not obsessed by the need for quick publications in high visibility journals. At that time, one could survive in the French scientific system without publishing for years. This gave me a unique opportunty to learn a lot, and develop many skills and ideas which were, much later, exploited in regular scientific work. On the whole, science was much more efficient at that time than it is now. It was still run by scientists. Another favourable circumstance in Luzzati's lab was its international character. There were in particular political refugees from Argentina and Poland , and a flux of visiting scientists from many countries abroad. There were also negative circumstances, which I was too naive to appreciate: The X-ray methodology was perhaps not fully mature, and there was a lack of expertise in the purification of nucleic acids.
Jean-Claude Monoulou and Hiroshi Fukuhara, from Piotr Slonimski's laboratory initiated me to the nucleic acid extraction and dosage procedures. I was finally able to prepare solutions of ribosomal RNA (or ribosomal RNA fragments). I studied ribosomal RNA in water and in ethylene glycol (which was thought to favour a stacked, single-stranded conformation) at several temperatures. The results are given in my thesis , pages 69-73. Looking at my thesis while writing this text, I discover that I had also made experiments on DNA, with or without bound magnesium (thesis, pages 74-75). This episode must have been erased from my memory almost immediately after the thesis' oral examination.
Fortunately, transfer RNA became commercially available, at least as bulk mixtures. So, I started doing small-angle X-ray scattering work on unfractionnated tRNA mixtures. Prior to this work, Witz had shown in his thesis (1964) that the tRNA molecule was more compact than an extended hairpin, that its radius of gyration (some measure of the elongation of the molecule) was around 25 Angströms, and he favoured a "boomerang" model. My own work lead to a slightly smaller radius of gyration, but otherwise did not produce anything striking beyond the work of Witz.
In parallel with the X-ray measurements, I started constructing models by hand. One of the motivations was to examine the plausibility of a model proposed by Willy Guschlbauer (Nature, volume 209, page 258). In this model, tRNA was almost entirely folded as a triple helix. So, I studied possible associations between nucleotide triads. I found that many triads were possible, but it was not easy to partition them into isosteric groups. From this work (made with archaïc, space-filling models, with wooden balls) I gained the conviction that all kinds of associations between nucleotides were possible.
Many things happened in 1965-1966. The genetic code was solved, and this opened new perspectives in molecular evolution. The first tRNA sequences were published, and the cloverleaf folding pattern gained credibility. And Crick published his wobble hypothesis.
This gave me plenty to think about. I had a (wrong) illumination towards the end of 1965. Everything that was not yet explained could be explained by assuming reverse transcription and reverse translation. So, I developped a theory of the origin of the genetic code (with a strong stereochemical component) which fortunately was never submitted for publication. I also developped the idea of reverse transcription for RNA viruses, but this was so obvious that I do not regret that it was not submitted. I also initiated computer work in molecular evolution, comparing haemoglobin and cytochrome c sequences. The work had perhaps some qualities, but it encountered strong resistance and was never submitted (see the forthcoming section on molecular evolution). To continue with half-aborted work, I also had a (correct) structural insight about RNA folding. It was about how, starting with a structure containing double-stranded sections, you could form triple-stranded sections. Briefly, a topological problem of strand switching could arise, and one could need an extra stem and loop region to make the switch. Such topological thinking was reflected in a subsequent paper on RNA topology ( - see forthcoming section on mathematical biology). Finally, in the context of Guschlbauer's three-stranded model for tRNA (then considered as just one of the functional conformations), there was the question of which association between nucleotides could be used on top of which association. I thus looked closely at succession rules between base-pairs in the stems of the cloverleaf, and found, for instance, a polarity in the GTUC stem: A GC pair, it seemed, had more chances of being followed by GC or UA than CG or AU.
In May 1967, while attending to a biophysical meeting in Paris, and listening to a talk by my friend Pierre Claverie, I had another illumination, but this time it was a profound one. It led me to the "missing triplet hypothesis" and from there to my subsequent carreer in the accuracy of molecular processes (see the forthcoming section on codon-anticodon recognition). After Claverie's talk, I raised a question about local geometry in RNA structures, which reflected my statistical observations on succession rules in tRNA stems. On this occasion, my name appeared printed for the first time in a scientific publication [P. Claverie, Journal de Chimie Physique, volume 65, year 1968, pages 57-65, my question being page 65]. The "missing triplet hypothesis" was received without enthousiasm in the laboratory. and it remained three years in my drawers. I was terrified at the idea that someone else might hit at the same idea, but it turned out to be an unfounded apprehension. The hypothesis was publlished four years later .
At the end of 1967 or perhaps the beginning of 1968, we had the visit of James Albert Lake. He had just completed a study of bulk yeast transfer RNA in solution, by small-angle X-ray scattering, in Beeman's Laboratory (Madison) The work was described in a paper in press in the prestigious Journal of Molecular Biology. So he had already completed the work which was still the remote goal of my thesis work.I had accomplished by then at least three full years of laboratory work, and did not have a single publication in sight.
The 1967-1968 period is extremely curious. On one side, I was involved in many extra-scientific activities, which, one might think, would have impaired my capacity for work. On the other hand, reading my 1968 report to the CNRS, it is clear that it was also the period during which the decisive part of my thesis work was carried out.
Outside science, I was very active in the labour union of scientific researchers, holding responsibilities in Gif, I was very involved, politically, in the movements of protests against the war in Viet-Nam, and I took a very active part in the May 1968 movement. In parallel, I met my future wife, and had my first son.
Scientifically, there was a complete technological overhaul. Marvelous space-filling atomic models (the Corey-Pauling-Koltun models) were bought, in sufficient amount to build, atom by atom, a complete tRNA model. Then, instead of using the theory of X-ray scattering by ellipsoids or cylinders, we could simulate a 3D model, atom by atom, and compute its theoretical X-ray scattering curve. The manual construction of models, and the generation of atomic coordinates corresponding to the model was entirely done by me (I always liked geometry). Concerning the theoretical X-ray scattering curve, there was at that time a problem of how to take into account the contribution of the surroundoing solvent. It happened that I had faciiities for algorithmic thinking, and I found a nice computational approach, that involved a partionning of the 3D space into cubes. Luzzati saw that it was physically valid, provided a corrective term was included, to remove the distorsions introduced by the sampling procedure. The method is still in use to-day (see the Section on Bio-informatics). I took inspiration from the cube partitionning method when, 15 years later, I proposed the method for fast comparison of nucleic acid sequences, widely used to-day, after being rechristened under the name of BLAST (see the Section on Bio-informatics).
The access to computers made another change possible. Until then, the originality of Luzzati's approach was in the use of an "infinite" linear collimation. The X-ray beam was assumed to extend laterally as far as needed. This infinite extension allowed the use of elegant mathematics to solve the "deconvolution problem" - in this case, the computation of the theoretical signal corresponding to point collimation, from the exprimental signal obtained with an infinite slit. Lake had devised a computer algorithm to solve the problem, which however did not always succeed, and I found a simple way to modify it to make it completely valid. The procedure is now famous in cognitive sciences under the name of "learning by retro-propagation" (See again the Section on Bio-informatics).
The major event, however, which changed the fate of my work was the encounter with Moshe Yaniv. Yaniv had been working in the laboratory of Sanger and Barrell in Cambridge on the valine-acceptor tRNAs from E. coli, which he purified from "tRNA fractions enriched in valine acceptor", a gift from Brian F. C. Clark. He offered us huge amounts, by the standards of the time, of pure tRNA Val, with which I could, at last, do decent X-ray work. We were, I think, in 1968, and I still did not have a single publication in sight.
Luzzati introduced several changes in the X-ray equipment. The X-ray acquisition line was made more reliable, the cells containg the samples to be analyzed were redesigned for work with smaller volume (pure tRNA was precious) and the slits were limited in length.
The gift of pure tRNA was, for my work, a fantastic piece of luck. Very soon, there was another lucky event, which however I did not exploit as well as I could. Alain Favre, who had joined biological research at the same time as I, and who was doing moecular spectroscopic work on RNA in Alan Michelson laboratory, irradiated the valine tRNA, and obtained a cross-link between two nucleotides far apart in the tRNA sequence: The partners of the cross-link were identified by Yaniv in Cambridge, so we had in our hands new, crucial information on tRNA 3D structure. (See the sections below on the tRNA model, and on the X-ray work).
I would like to mention a few people who played a role in my beginnings, and whom I did not yet cite. First, Pedro Saludjian from Argentina who had a broad culture, and a capacity to bring interesting insights on almost any topic. He contributed in a very stimulating manner to the scientific discussions, and possibly in the upgrading of the X-ray methodology. On the other hand, after the untimely death of his wife, he had increasing difficulties in transforming his work into straight scientific publications. He would not have survived in to-day's scientific system, yet his presence was at least as essential as that of others who were doing the ant's work. Outside the laboratory, but at the centre of the communitee of refugees from Argentina, was Luisa Hirshbein, who helped us establishing useful contacts with colleagues from other disciplines. Luzzati's group was one among the numerous teams of the "Centre de Génétique Moléculaire", headed simultaneously or alternatively by Boris Ephrussi, and his former junior associate, Piotr Slonimski. Piotr had a personality with many facets, and a great talent for manipulating people. I remember participating to a demonstration with him, shoulder to shoulder, during the May 68 movement. Otherwise, our relationship was a conflictual. When the cytochrome c sequences determined in the laboratory of Margoliash started to accumulate, Slonimski teamed with a Polish mathematician, Krszyvicki, on the computer analysis of protein sequences. They were not pleased (perhaps rightfully) with my own computer work on sequences, and I was kept at a distance. Much later, in 1991, Slonimski helped me in my carreer, putting his weight at the CNRS for a promotion to my present position (I had been blocked nearly ten years by Marianne Grunberg-Manago and Jean-Pierre Ebel). Boris Ephrussi was much older than Slonimski. People thought he was distant and were afraid of him. But he took me in sympathy, he often invited me in his office for a chat, and made many useful scientific suggestions in connection with my interests not directly related to the thesis. As a tribute to him, I later dedicated one of my best papers (in 1979, on visual illusions, in J. Theoret. Biol.79, 167-201) to his memory.
I performed my military duties in oct. 69 - august 70 as a second lieutenant in the transmission corp. I had plenty of free time which I used writing my thesis and writing (in secret) the "missing triplet hypothesis" which was subsequently submitted to the Journal of Molecular Biology. Upon my return, Luzzati requested profound changes in the thesis, so I spent time rewriting the thesis, and also writing an article on the topoplogy of nucleic acids. Luzzati and I wrote for the Journal of Molecular Biology the article describing the small-angle X-ray scattering work. By that time, I had grown disenchanted with "lock and key" explanations and the "structure-function" paradigm, and I was extending my ideas on molecular recogntion to protein- protein interactions. I attempted to build an autonomous team with two biophysicists of Luzzati's laboratory, Claude Reiss and Pedro Saludjian, on the theme of "protein-protein interactions", but we were immediately stopped.
My main interest, at that time, was the origin of the genetic code (see the forthcoming section on the origins of life). In the last days of April 71, Boris Ephrussi told me about an international meeting which would soon be held in the palace of Versailles "De la Physique Théorique à la Biologie" (From Theoretical Physics to Biology). He said that he knew the organizer, Maurice Marois, and he would support my application. I was taken at once. It turned out that the meeting was a big meeting with top scientists from many disciplines, and an incredibly large number of Nobel prize winners or future Nobel prize winners.
I was perhaps the only young and obscure participant in the audience. This meeting had two important consequences for my future carreer. One talk made a very deep impression on me, it was by Bela Julesz, on random-dot stereograms (see forthcoming section on stereoscopic vision). And there, I encountered Leslie Orgel. I had in fact just one or two brief contacts with him. Towards the end of the meeting I gave him a reprint of "the missing triplet hypothesis" and he then looked at me seriously. He offered me to make a post-doctoral period in his laboratory at The Salk Institute. I agreed on the principle, and started to reorganize my life accordingly, switching from Luzzati to Chapeville's laboratory (see forthcoming sections on codon-anticodon recogntion, and on the origins of life).
MOLECULAR MODEL FOR TRANSFER RNA [7, 9]
In the 1964-1969 period, we were convinced that many, many years would elapse before a tRNA structure could be derived from crystallographis studies. It is under this illusion that I spent so much of my early years in science in an attempt to construct a molecular model for tRNA. Furthermore, I was strongly influenced by the "structure-function" paradigm, according to which the ultimate explanation of the function would be found in the structure. Accordingly, I expected that the tRNA structure would provide the key to the nature and the origin of the genetic code.
Having in hands my experimental results on small-angle X-ray scattering by tRNA, and the original, yet unpublished result by Favre, Michelson and Yaniv on the spatial proximity between two residues far apart in the linear sequence, I produced a model which was "not too bad", but differed substantially from the now accepted L model.
I am entirely responsible for the mistake, and present here my apologies to Favre and Yaniv for spoiling this opportunity of reaching fame.
In 3D, our model  resembled the crystallographic L model more than any other tRNA model published before, or after this one (except the L). The mistake concerns the two base-paired segments of the acceptor stem and the GTUC stem. In the L model, they are co-stacked at a right angle to the rest of the molecule. In our model, they are folded back around their hinge and parallel to the main axis of the molecule. The extra-loop was folded around the DiHU stem and, in a subsequent review , placed even more correctly in the large groove of the DiHU stem. Furthermore, I suggested that single-stranded regions could interact by making purine sandwiches "the residue 9 (frequently a purine in tRNA sequences) may be stacked between residues 44 and 45 of the extra arm" - a proposal  which turned out to be correct, and which was ahead of its time.
Favre and Yaniv were aware of all the work done on chemical modification of tRNAs, from which people were deriving a map of the residues which were exposed to the solvent, and those which were base-paired to other nucleotides. I did not pay too much attention to this type of data, and considered that they did not provide first quality clues.
On the other hand, I relied heavily on clues indicating that all the stems of the tRNA were parallel: There were preliminary reports from crystallographic studies claiming the parallelism of all stems, and I had also heard from Russian biophysisicists that all tRNA stems were parallel according to some "magnetodynamic" (?) method unknown to me. Furthermore James Albert lake had shown, in his J. Mol. Biol. 1968 article that if the tRNA molecule was composed of two separated big stacks, this would generate a shoulder in the X-ray scattering curve, which I never saw in my results. I extrapolated from this that similarly, an L model would produce a shoulder, and was therefore excluded.
In March 1969, Favre, Yaniv and I attended an EMBO meeting in Cambridge, devoted to transfer RNA. On this occasion, we paid a visit to Francis Crick. He said that he had heard from Luzzati that the tRNA molecule was long and thin, and he was thinking in terms of a model in which the the amino acid stem was stacked over the GTUC stem (as it is), in which the two other stems would also form a stack (as they do) and in which the two big stacks would interact through the DiHU and the GTUC loops (again, correct). I do not remember whether or not he saw the two big stacks forming a corner. He probably did. I gave my arguments in favour of my model, and he gave us his blessing. The tRNA meeting was an informal one, in the original tradition of EMBO meetings. An introductory talk was perhaps scheduled in each session, but after that, all the time was available for free discussions. Friedrich Cramer introduced the first session, presenting his model, mainly based on nucleotide accessibility work, in which three stems were closely packed in parallel, with their loops interacting, and the acceptor stem extending in the opposite direction. He insisted that his model was in agreement with every piece of available experimental evidence. After his talk, I bounced from my chair, and presented, with youthfull arrogance, my own view of tRNA structure. Among the people attending the meeting there was a young Cambridge researcher, Michael Levitt. He showed great interest in my work, and spent considerable time discussing with me all the evidence I had. We have maintained good ties from this period.
After the Cambridge meeting, we wrote, for "Nature" the article describing the model and with Sanger's support, it was published without difficulty. In the acknowledgements, we thanked, among others, Brian F. C. Clark "for the gift of tRNA fractions enriched in valine acceptor". Two months later, Michael Levitt published in Nature his own model, in which he embodied his original observations on correlated changes in tRNA sequences. He had observed a correlation between two remote residues, which could have been accomodated, with minor changes, in our model. But he proposed a radically different model in which more attention was paid to nucleotide accessibility, and less importance was given to the radius of gyration (too small, in his model, most the electronic mass being at the centre).His model became at once the most popular one, and it has been quoted 260 times whereas ours was quoted 57 times. This tells us more about citation idiosyncrasies, than about the respective merits of the two models.
In 71 and 72 I attended several meetings on transfer RNA. Then, In Sept. 72 I moved to the Salk Institute, close to San Diego. I was still believing in my model, and in November 72, I gave a seminar on my tRNA work (invited by John carbon) at the University of California, Santa Barbara. .
In January 1973, Sung-Hou Kim and co-workers published in Science (179, 285-288) the L model, derived from their crystallographic work.
In May-June 1977, I was spending two months in Rudolf Rigler's laboratory, in Stockholm. I learnt that the atomic coordinates of tRNA from the crystal structures were available there, so I computed the theoretical small-angle scattering curve from the model. It turned out to fit my data even more closely than the model I had proposed. I wrote a letter to Alex Rich with a figure showing the experimental and the theoretical curves, suggesting to him to insert the figure in his next review on tRNA, but he declined the invitation, saying that I should make a note of my own on this topic.
I derived from this episode in my life a big lesson, and I believe that I never repeated the same type of mistake again in my whole carreer. But where was the mistake? It was definitely not where most people would think. People can say that I did not pay enough attention to the nucleotide accessibility work. But the fact is that, just after the Cambridge meeting, and the discussion with Crick, I did construct the L model. However, I did not compute the theoretical X-ray scattering curve for this model, being (wrongly) convinced that the theoretical curve would display that famous shoulder which was never observed in the data. So, I have now a principle of exploring any line of evidence in a completely systematic manner, even though there may be obvious, immediate objections to some of the variants.
SMALL-ANGLE X-RAY SCATTERING WORK 
It seems now unreal to me that I spent so much time collecting data on the diffusion of X-rays at small angles by solutions of transfer RNA. The experiments were tedious. It took about 12 hours to collect the scattering curve from a sample. Towards the very small-angles, the signal was strong, but one neeeded to extrapolate to zero concentration of solute, and this was highly perilous. At wide angles, the signal was very weak. I was collecting data much beyond the angular region explored in other works, and in which the magnitude of the signal had fallen to less than 1% of its value close to the origin. One had to be very cautious in drawing a regular plot from the highly noisy records at large angles. I guess that to-day, with the availability of powerful X-ray sources, all my results can be obtained much more easily (and perhaps they have, but I no longer follow this line of work).
In any event, I published, with Yaniv and Luzzati the small-angle X-ray scattering curves for several tRNA states:
- State 1: the standard, deacylated tRNA Val (in fact, a mixture of two isoacceptors from E. coli)
- State 2; the deacylated tRNA Val which had been cross-linked after UV irradiation
- State 3: an acylated form of tRNA Val, which was further modified with a bulky group (phenoxyacetyl-) attached to the amino acid. This modification was in fact the trick (invented by Sanger ?) to isolate the purified tRNA.
- plus some controls
The study showed that all three states had nearly identical shape parameters. On the other hand, there were differences were they were not expected, in the "large-angle" region. We could not find any straight explanation for these differences, and no plausible source of artefact. In dispair, we wrote in the abstract: "it is suggested that upon fixation of the amino acid the tRNA molecule undergoes small structure modifications, which affect the distribution of the charged groups at the outer surface of the molecule; as a consequence the solute-solvent interactions, and probably the binding of magnesium, are altered." The observations were highly reproducible. But, as Leslie Orgel once said to me "mothing is more reproducible than an artefact". So, I leave it to others to corroborate or refute our observations.
Apparently, the most useful part of the work, to-day, is the "cube" method for calculating the X-ray scattering in the presence of solvent.
I also devoted some time looking at the make-up of tRNA stems: which base-pair was followed by which, which were the most common neighbours on both sides of a G.U pair, in what context did one find non-wobble pairs such as U.U ? One implicit assumption in this work was that the frequencies of occurrence would reflect the energetic preferences. From there, one could hope to bring some light on various topics, and I was mostly thinking of codon-anticodon recognition errors.
Using the available sequences, I was doing all the statistics by hand. I found some very significant deviations from randomness in the nearest-neighbour frequencies, and wrote a note for Nature, which was rejected, then for the European Journal of Biochemistry. The paper was entitled: "Statistical study of the base-paired regions in transfer RNA and their relevance to the structure of 5s RNA". The reviews were lukewarm but not negative, and revision was encouraged. The Editor wrote (October 21th, 1971): "We would appreciate, therefore, if you could prepare a revised version to meet all the criticisms of the referees". For reasons I have forgotten, I did not submit a revised version. Possibly, having, in the meantime obtained an invitation to make a review for Progress in Nucleic Acid Research, I considered that it was just as well to incorporate into this review the gist of my observations on base-pair statistics .
Much later, when I worked with Jean-Pierre Dumas on the prediction of RNA structure, we studied the base-pair statistics and attempted to develop an energy-model that would reflect these statistics.
To-day, I am convinced that the nearest-neighbour statistics are shaped by many factors, so they are not good predictors of energy preferences. A most important factor, in eukaryotic organisms, is the mutability of the CpG sequence in DNA, leading to its under-representation. In my experience, energetic guesses derived from statistical frequencies never worked well (see also the forthcoming section on codon-anticodon recognition). But it is also the case that energy values derived from biophysical measurements did not perform better, in the prediction of RNA structure. This is due to the fact the energy values express differences between two states; and the proper reference states to take into account in the folding simulations are not necessarily the same as those which are implied in the biophysical determinations .
REFERENCES AND NOTES
 Ninio, J. (2002) Maurice Hofnung - quelques souvenirs. Research in Microbiology 153, 481-484.
 Ninio, J. (1991) La biologie buissonnière. Seuil, Paris.
 Ninio, J. (1971) Etude de la structure de l'ARN de transfert par diffusion centrale des rayons X, et de ses implications biologiques. Thèse d'Etat, Université Paris 7.
 Ninio, J. (1971) Properties of nucleic acid representations. 1. Topology. Biochimie 53, 485-494.
 Ninio, J. (1971) Codon-anticodon recognition : the missing triplet hypothesis. J. Mol. Biol. 56, 63-82. .
 That one could be, in the 1960's, simultaneously effcicient in science and politically active seems very strange to-day. The explanation of the paradox ls in the fact that scientific research (and all kind of creative work) was in fact much more efficient in these days than it is in our days of economic and managerial cretinism (see forthcoming Sections on Economics or Science and Society).
 Ninio, J., Favre, A. et Yaniv, M. (1969) Molecular model for transfer RNA. Nature 233, 1333-1335.
 Ninio, J., Luzzati, V. and Yaniv, M. (1972) Comparative small-angle X-ray scattering studies on unacylated, acylated and cross-linked Escherichia coli transfer RNA Val. 1. J. Mol. Biol. 71, 217-229.
 Ninio, J. (1973) Recognition in nucelic acids and the anticodon families. Progress in Nucleic Acid Res. Mol. Biol. 13, 301-337.
 Ninio, J. (1979) Prediction of pairing schemes in RNA molecules. Loop contributions and energy of wobble and non-wobble pairs. Biochimie 61, 1133-1150.