In the beginning of the 1970s, Ben Hesper and I started to use the term “bioinformatics” for the research we wanted to do, defining it as “the study of informatic processes in biotic systems”. (Although several public sources [see below] trace the origin of the term to publications by us that appeared in 1978 
, in fact we were using it as early as 1970, proposing the definition above in an article in Dutch that is not generally accessible 
It seemed to us that one of the defining properties of life was information processing in its various forms, e.g., information accumulation during evolution, information transmission from DNA to intra- and intercellular processes, and the interpretation of such information at multiple levels. At a minimum, we felt that that information processing could serve as a useful metaphor for understanding living systems. We therefore thought that in addition to biophysics and biochemistry, it was useful to distinguish bioinformatics as a research field (or what we termed a “work concept”).
Indeed, at the birth of molecular biology it was recognized that a central research theme should be how living systems gather, process, store, and use information 
. This focus on concepts related to information is, for example, reflected in the terminology “genetic code”, the central dogma as the unidirectional flow of information, etc. A nice monograph entitled “From Deoxyribonucleic Acid to Protein: Transfer of Genetic Information” 
summarized the state of the art in molecular biology before the “sequence age”, unraveling for me the essential processes that, at the time in genetics undergraduate texts, were buried in “bead genetics”. It seems that recently, after a dormant phase, such information-centric terminology has become more prevalent again (e.g., in terms of identifying a distinct research field 
and focusing on such processes as sensing the environment 
and dynamic phosphorylation and methylation codes 
We were embedded then within theoretical biology. At the time, after general systems theory 
had come and gone, theoretical biology was in a mild resurgence in acceptance. The series of books entitled “Towards a Theoretical Biology”, edited by Waddington 
(reprints of which are underway), had appeared a few years earlier. In 1972, the main topic at a meeting organized by BSRC (Biological Science Research Council) Developmental Biology in collaboration with the Society for Experimental Biology was mathematical models of development.
Stuart Kaufman was there, presenting his work on random Boolean networks, which introduced the concept of large-scale transcription regulation networks and viewed a cell type as an attractor in a multidimensional dynamical system 
. It is striking that in the year 2000, Huang and Ingber reintroduced these concepts to the experimental molecular biology community 
and later beautifully illustrated their power by demonstrating alternative trajectories to neutrophil differentiation on the basis of temporal gene expression data of 2,773 genes 
At this same meeting, models and experiments in such areas as oscillatory enzyme dynamics (e.g., 
), positional information 
, and bi-stability in gene regulation 
were presented and hotly discussed. Spatial pattern formation was one of the central topics, contrasting Turing systems 
with gradient-based systems 
. Francis Crick, who in that period published some papers on gradients in development 
, attended the meeting. Skeptical about the emphasis Turing Patterns were (still) receiving, Crick quoted Turing as saying in reaction to enthusiasm about his work: “Well, the stripes are easy but what about the horse part?” To go “for the horse part”, i.e., to go beyond pattern formation to multilevel models of development and morphogenesis, became one of the long-term goals of our nascent work concept “bioinformatics”.
Also at about that time, John Maynard Smith gave a lecture in Utrecht and posed a similar challenge with respect to evolutionary biology as Turing's challenge relative to developmental biology. While evolutionary models mainly dealt with invasion of mutants and changing allele frequencies, the question of how evolution leads to complex organisms was not addressed. As Maynard Smith expressed it: “As good evolutionary biologists we should go once a year to the zoo and visit the elephant. We should greet it and say ‘Elephant, I believe you got about by random mutation’”. To meet the challenge of a “constructive evolutionary biology” became another long-term goal of bioinformatics as we envisioned it.
Research in artificial intelligence at this time was exploring new representations of information processing systems, often inspired by biological systems, e.g. neural network models for learning and pattern recognition 
, genetic algorithms 
for optimization, “actors” 
for semi-independent parallel processing, and “turtle geometry” 
, demonstrating the power of an individual self-centered approach to generating and/or understanding more global structures.
We felt that the re-introduction of biologically inspired computational ideas back into biology was needed in order to begin to understand biological systems as information processing systems. In particular, a focus on local interaction leading to emergent phenomena at multiple scales seemed to be missing in most biological models.
At the time, molecular biology was of course not a heavily “data-driven” science, as it would become with the advent of massive sequencing projects. Indeed, data-driven science was looked down upon, both in molecular biology and in theoretical biology. However, data-driven research was being done in the more traditional parts of biology, ecology, and taxonomy. I had just finished a data collection survey on water plant vegetation in India, Czechoslovakia, and The Netherlands and had become dissatisfied by the local state of the art of data processing, which comprised shuffling large tables by hand. At the same time, pattern recognition methods had already been introduced as “numerical taxonomy” 
, as well as in ecology 
. Although modeling and pattern analysis were (and still often are) seen as separate endeavors, we felt that for bioinformatic research they were both needed and should be combined: first, to analyze patterns of variation at multiple levels in organisms; second, to detect emergent phenomena in models; third, to compare the outcome of such models with “real” data; and finally, and most profoundly, because the relationship between genotype, phenotype, behavior, and environment itself can be seen as a type of pattern recognition or pattern transformation 
, and understanding these processes was the core of bioinformatic research.
In short, under the heading of bioinformatics we wanted to combine pattern analysis and dynamic modeling and apply them to the challenge of unraveling pattern generation and informatic processes in biotic systems at multiple scales.