Computer modeling, genomics, and prostate cancer prognosis


A newly published paper on the development of prostate cancer in patients of ≤ 55 years of age claims to have shown that

Using a newly-developed computer model, it is now possible to predict the course of the disease in individual patients.

The new paper by Gerhauser et al. in the journal Cancer Cell (see also a supporting announcement about the findings on the Science Daily web site) reports that the German research team first carried out a detailed analysis of data from 292 prostate cancer genomes to identify a series of “age-related genomic alterations and a clock-like enzymatic-driven mutational process” that contribute to the earliest mutations in prostate cancer patients.

In so far as your sitemaster is capable of understanding this paper and related information, it appears that the initial data analysis has permitted

  • Appreciation of the mutations that must occur occur first, to start the process of transformation of prostate cells into cancer cells
  • Greater understanding of the consequent order in which additional genetic mutations must take place over time for differing types of prostate cancer to develop
  • Identification of four molecular subtypes of prostate cancer that progress at differing rates (and which, by implication, may be need to be managed in differing ways)

The research team then used these results to develop a sophisticated, computer-based modeling system that seems to be capable of predicting the likely course of the disease in individual patients. A “graphical abstract” of the prognostic system is shown below:

According to one of the authors:

Our team is currently busy incorporating our computer model into the treatment process [at the Charité – Universitätsmedizin Berlin]. This will enable clinicians to model a particular treatment’s likelihood of success. As for the timescale involved, we expect it will take two to three years for this algorithm-based method to become clinical routine.

Your sitemaster wishes to be clear that he has only seen the abstract of this new paper by Gerhauser et al. and the associated media statement (and even if he had seen the whole paper, he doubts that he has the technical knowledge to understand the most of the details).

On the other hand, what your sitemaster is very aware of is that there has been an enormous amount of research going on into the ways in which we can use genetic and related clinical data to be able to build this type of prognostic computer-based modeling system for all sorts of different types of disorder, so he is not surprised by this development. The questions that are going to be most important in taking the next steps are going to include the following:

  • Just how accurate are the prognoses that can be made using this system (and other similar systems in the future)?
  • What are the costs associated with the use of such systems, and can we actually afford to use such systems in the day to day practice of patient care?
  • Do systems like this only work in prostate cancer when the patients are diagnosed at a young enough age? (Note that this system is based on information from a cohort of men who were all diagnosed at ≤ 55 years.)
  • To what extent will we be able to correlate the prognostic data available from such systems with high quality clinical outcomes data to be able to predict the most appropriate types of management for individual patients?

The media information on the Science Daily web site (mentioned previously) indicates that

In an effort to improve the reliability of prognoses, the research consortium is planning to spend the next few years collating additional data on thousands of patients, which they will then use to further develop and enhance their computer model. They will achieve this by working with Berlin’s newly established urology network (Hauptstadt-Urologie-Netzwerk), which brings together urology specialists from Charité and private practice. Their ultimate aim is to make it easier for physicians to decide on the most suitable treatments for individual patients.

In other words, it is going to take some time to convert the initial promise of the research reported into a practical tool for clinicians, and the degree to which that may prove to be useful will then need to be carefully tested over time.

The actual implementation of truly personalized medicine is going to be highly dependent on the evolution of these types of computer-based prognostic modeling and the associated therapeutic interventions and outcomes. Your sitemaster expects that we are going to find all sorts of “hiccups” along the way, but if it was possible to use such methods simply to be able to project, with accuracy, which patients had clinically significant prostate cancer and which ones did not with a relatively high degree of accuracy (say 80 percent or higher) at initial diagnosis, that in and of itself would be a major step forward.

2 Responses

  1. We already have risk-based treatment decision systems for prostate cancer. It is no surprise that oncogenes and tumor suppressor genes like PTEN and RB1 and TP53 may have prognostic significance. The team is going to make new studies as to the external validity. However, the clinical challenge is the extent to which the new model will improve the risk categorization compared with the traditional model based on Gleason score, clinical stage, PSA level, and PSA kinetics; whether the change in risk categorization might change treatment decisions; and whether the change in treatment would improve overall survival.

  2. Dear Finn:

    One of the great benefits of computer-based models of this type is that it should be very easy indeed to integrate the clinical data and the genomic data to see how much difference the addition of the genomic data may make (first with regard to the selection of therapy and then, subsequently, with regard to actual outcomes over time).

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out /  Change )

Google photo

You are commenting using your Google account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s

This site uses Akismet to reduce spam. Learn how your comment data is processed.

%d bloggers like this: