Some of the most exciting science presented at the 2011 NCRI conference on its final morning was connected by a single theme: stratified or personalised oncology. This combines two well known ideas: that of the need to provide the right drug for the right patient at the right time, and that of the classification of any tumour type into perhaps many multiple sub-classes.

Firstly, plenary speaker Sabine Tejpar from University Hospital Leuven in Belgium described a picture of colorectal cancer as a heterogeneous disease where stratification is necessary if each patient is to be given the most appropriate treatment. However, she could only describe current progress towards identifying those variants and the appropriate treatments as "steady but extremely slow". The heterogeneity of this condition starts with tumour location. Rectal cancer is commonly considered a different disease, but there are also important differences in both the biology of a tumour and its prognosis depending on whether it develops in the right (ascending) or left (descending) colon. The mean quoted survival time for tumours on the left side is almost twice that for those on the right.

Most of Tejpar's talk, however, was devoted to the genetic heterogeneity of colon cancer. Prognostic mutations have been well characterised in this disease in only one gene, KRAS. Another gene in the same pathway, BRAF, is also mutated in many cases of colon cancer. She described gene expression profiles taken from colon cancers that had been divided into three categories, KRAS mutant, BRAF mutant and double wild type. Interestingly, a substantial minority of both the KRAS mutant tumours and the double wild type ones had gene expression profiles that were "BRAF mutant-like". More recent work has divided colon cancers into four sub-groups based on their patterns of DNA methylation. Tejpar summarised her talk by predicting that these tumours would eventually be classified into maybe seven or eight "clean" subgroups based mainly on genetic features, and that these would predict both prognosis and drug response.

Drug developers can only make the best use of stratification in clinical trials if biomarkers are incorporated in the trial design. Caroline Dive from the Paterson Institute for Cancer Research, Manchester, UK, introduced an interesting session on biomarker validation and use in adaptive trial design. An adaptive clinical trial is one where the design makes provision for the protocol to be changed during the trial. Howard Scher from Memorial Sloan-Kettering Cancer Center, New York, USA described the use of biomarkers in clinical trials of drugs for prostate cancer, including Medivation's promising new androgen receptor antagonist, MDV3100. Jose Tabernero from Vall d'Hebron University Hospital, Barcelona, Spain explained how his group uses biomarker defined endpoints in Phase I to set a wider and potentially more useful dose range for Phase II trials. Lucinda Billingham from the University of Birmingham, UK, described the BATTLE lung cancer trial, a complex trial of four treatments in five biomarker-defined groups that used Bayesian statistics in its design. In this trial, early results were used in randomising later patients, who had a higher chance of receiving a treatment that was yielding good results in their particular cancer subtype.

The conference ended with a tour de force of a plenary lecture, delivered to a packed auditorium by Michael Stratton of the Wellcome Trust Sanger Institute, Cambridge, UK. He used DNA sequence analysis and comparison of the complete exomes (protein coding genomes) of 69 breast tumours to explain some of the latest thinking in the evolution of the cancer genome. Although an individual cancer typically contains a great many somatic mutations, the majority of these are "passenger" mutations; only a few in each tumour are identified as the "drivers" of carcinogenesis. About two percent of the maybe 21,000 genes in the human genome have so far been identified as drivers of carcinogenesis in at least one tumour type.

Stratton's comparison of breast cancer exomes led to the identification of seven more such genes, five of which were thought to be inactivating (recessive) mutations and two activating (dominant) ones. The tumours were very heterogeneous in the number and identity of genes containing driver mutations; the presence of a few tumours bearing none led him to speculate that some tumours might be driven by DNA methylation changes alone. After commenting on the mechanism of mutation and the evolution of tumour genomes, he ended by predicting that the International Cancer Genome Consortium's project to sequence some 25,000 tumour genomes in the next 5-7 years would lead to the elucidation of a "complete atlas" of genetic changes in cancer. The field of personalised oncology should become even more interesting.

It will be interesting to see how the field progresses even in the next year. NCRI chair Dame Janet Husband closed the meeting by inviting delegates to the eighth NCRI cancer conference, to be held at the same venue from November 4-7, 2012. See you there!