The concept of personalised oncology is often compared to weather prediction. The idea being that with increased amounts of data from patients (genetic sequencing, phenotypic characterisation of cancer cells, imaging etc.) and more advanced mathematical models, we will be able predict tumour progression and responses to therapy in individual patients with increased accuracy.
When making this comparison, the patient data is analogous to the current state of the atmosphere (temperature, air pressure, wind speed and direction etc.) used as input to fluid dynamics models, which can for example be used in order to predict the future course of a hurricane (or in terms of cancer predict the rate of growth under different therapies).
The current state of personalised medicine is however falling short on both accounts: the data acquired from patients is meager (although microarray data is 'big' it's pretty useless as input to computational models), and our present-day theoretical understanding of tumour growth is limited. Even if we had all the data we could dream of it would most likely be useless because of our ignorance.
Although the analogy seems to fall short it actually sheds some light on our inability to predict tumour growth, the reason being that meteorology was in a similar situation roughly a century ago. Before the advent of digital computers, weather prediction was a difficult business based on previous experience and certain rules of thumb. Or in the words of Lew Fry Richardson in the preface of "Weather prediction by numerical process" (1922):
The process of forecasting, which has been carried on in London for many years, maybe typified by one of its latest developments, namely Col. E. Gold's Index of Weather Maps. It would be difficult to imagine anything more immediately practical. The observing stations telegraph the elements of present weather. At the head office these particulars are set in their places upon a large-scale map. The index then enables the forecaster to find a number of previous maps which resemble the present one. The forecast is based on the supposition that what the atmosphere did then, it will do again now. There is no troublesome calculation, with its possibilities of theoretical or arithmetical error. The past history of the atmosphere is used, so to speak, as a full-scale working model of its present self.
This process of manually predicting the weather is similar to the way clinicians in the present day decide upon different cancer therapies. Treatment choices are based on sparse data and previous experience of similar patients.
The seminal work of Vilhelm Bjerknes and the above quoted book lay the foundation of quantitative weather prediction. What was still lacking was however the computational power, which held back large scale weather forecasting by another 30 years. The situation in mathematical oncology today is however reversed. We have all the computational power we need, but still lack the appropriate theoretical understanding. Hope fully one day that will change.