Blood tests have proved to be a promising tool for detecting and monitoring cancer. Researchers at Chalmers University of Technology and the University of Gothenburg, Sweden, have now developed a new method that makes it possible to analyse samples containing as little as 5 per cent cancer DNA in the blood, compared with the 15–20 per cent required today. This method could lead to better cancer care and improved monitoring of tumour progression.

Analysing changes in tumour DNA using blood tests is a technique that is being explored in several clinical trials worldwide. Current analytical methods work well when the amount of cancer DNA is relatively high, at around 15–20 per cent of the total DNA in the blood. However, the level of cancer DNA is often considerably lower than this. This may mean that the sample quality is not good enough for detailed analysis.

‘We wanted to develop a method that works particularly well in difficult cases where there is very little cancer DNA in the blood and a lot of what we consider noise – that is, mainly healthy DNA. Our results show that the new method performs better with samples involving low levels of cancer DNA, where the proportion is around 5 per cent. So, it works exactly as we had hoped,’ says Lotta Eriksson, a doctoral student in the Department of Mathematical Sciences at Chalmers and the University of Gothenburg.

Blood-based methods being tested in clinical trials are often used to determine whether cancer can be detected at all. It is difficult to obtain a more detailed picture, partly due to high costs and poor sample quality.

The new method, BayesCNA, can extract information that was previously hidden in low-quality samples and provide more detail about the tumour’s composition. This can help to provide a better understanding of how a patient’s cancer changes over time.

‘When the treatment is effective, the amount of cancer DNA in the blood drops significantly. This makes it more difficult both to detect the cancer and to monitor how it changes. It is important to be able to analyse samples containing low levels of cancer DNA to gain a clearer picture of how a patient responds to treatment,’ says Eszter Lakatos, Assistant Professor in the Department of Mathematical Sciences at Chalmers and the University of Gothenburg.

At present, a tissue sample from the tumour itself is required to obtain detailed information about its composition. The ability to monitor tumour progression using blood tests could lead to significantly better care for cancer patients.

‘A patient may undergo surgery once or twice, whereas blood tests may be taken at intervals of just a few weeks during treatment. If we can obtain information about tumour changes from the samples, we can monitor developments much more closely and see what happens between treatment sessions. This can help doctors make more informed decisions, such as tailoring treatment to the tumour’s composition,’ says Eszter Lakatos.

The method has been developed to analyse data from what is known as low-pass whole-genome sequencing, a technique that provides a general overview of the DNA structure. The technique has major financial benefits but provides limited information because the data quality is poor.

‘You could compare it to skimming through a book rather than reading it properly. We get an overview of the DNA structure, but not a detailed picture,’ says Eszter Lakatos.

The new analysis method uses a statistical algorithm to amplify the very weak signals present in this type of sample.

‘Nowadays, machine learning is used to solve a great many problems, and we tried those methods first. But, to our surprise, it turned out that classical statistics worked better in this case, which was particularly pleasing to us mathematicians and statisticians,’ says Lotta Eriksson.

The next step is to analyse the information the method provides on tumour composition. The researchers are keen to develop a further method for identifying the hidden characteristics of the cancer that influence how patients respond to treatment.

‘If we can demonstrate that this information is useful, we hope it will lead to more collaborations and wider adoption of our method within the research community. In the long term, I hope that the methods we develop can be used in clinical trials and, with any luck, make a difference to the care of cancer patients,’ says Eszter Lakatos.

The article ‘Sensitive detection of copy number alterations in low-pass liquid biopsy sequencing data’ has been published in the journal Briefings in Bioinformatics. The authors of the study are Lotta Eriksson and Eszter Lakatos, both of whom are currently in the Department of Mathematical Sciences at Chalmers University of Technology and the University of Gothenburg, Sweden

The study was funded by the Swedish Research Council and Chalmers’ Health Engineering Area of Advance.