Deep Learning Algorithms Are Expected To Boost Liquid Biopsy Techniques

By | May 27, 2019

Liquid biopsies are cost-effective and reduce the sress of the biopsy procedure   


According to Wikipedia, the earliest diagnostic biopsy was conducted by an Arab physician in the 11th century. The term “biopsy” derives from the Greek word bios which means life and the also Greek word “opsis” which means sight. It was first introduced in the 19th century and since then medical researchers and practitioners have developed several techniques to perform biopsies on living organs or tissues needed further observation. Modern instruments such as intestinal biopsy tubes can extract samples from parts of the body where it is difficult to have access or even hazardous. 

In the cases described above, this histologic procedure is invasive and quite complicated putting a financial burden, not to mention the extra strain on the patient. Liquid biopsy is an emerging and very promising technique to perform biopsies. It detects and isolates the circulating tumor DNA and exosomes as a source of genomic and proteomic information in patients with cancer. As every new technique faces technical hurdles which affect effectiveness and accuracy. New technological developments and next-generation sequencing analysis allow a broad application of liquid biopsy in a wide range of settings. Initially correlated to prognosis, liquid biopsy data are now being studied for cancer diagnosis, including screenings, and most importantly for the prediction of response or resistance to given treatments. In particular, the identification of specific mutations in target genes can be detected more easily as liquid biopsies are easier to be taken. That translates to more data for the physicians to analyze in real time with a specific treatment and decide on the appropriateness of treatment or the identification of secondary resistance, aiming to diagnose disease progression early enough.

Since the first description of tumor cells in blood circulation in 1869, substantial progress has been made from biotechnological applications to isolate CTCs from heterogeneous blood components. However, the systematic application of liquid biopsy in real practice is still hindered by hurdles, such as unsatisfactory specificity and sensitivity, lack of standardization, and high economic and human resource costs, and still offers many challenges. In fact, due to the low concentration of CTCs, ctDNA, and EXOs currently recoverable from the patient, the analytical results sometimes suffer from unsatisfactory specificity and sensitivity.

We are already at the second stage of liquid biopsy methods

The initial methods applied for liquid biopsies focused on either ctDNA or protein. But as Nickolas Papadopoulos, Ph.D., Professor of Oncology at the Johns Hopkins Kimmel Cancer Center explained in an article published in Technology Networks “The amount of ctDNA in the blood in early-stage cancer is very low and can constitute only 0.01% of the total cell-free DNA.” This can create stochastic events based on the amount of blood used for the test. In addition, many cancers do not ‘shed’ their tumor DNA into the circulation so some cancers would not be detected if only the ctDNA factor was assayed and the test, though specific, would not be sufficiently sensitive.

The first report of dual ctDNA and protein liquid biopsy research that combined blood tests for KRAS gene mutations with protein biomarkers to determine whether the combination of these markers was superior to any single marker, reports that the comparison between  patients with pancreatic ductal adenocarcinoma (PDAC) and healthy people led to a sensitivity of 64% with a specificity of 99.5%. This is far better from any test that analyzes one of the two factors.

Promising two-factor liquid biopsy with less than $ 500, CancerSEEK is a promising solution for the liquid biopsy market


Then a second report focused on other early-stage cancers, including ovarian, liver, stomach, pancreatic, esophageal, colorectal, lung, and breast. The test, named CancerSEEK demonstrated a median sensitivity of 70% (lowest for breast cancer at 33%, highest for ovarian cancer at 98%) with a specificity higher than 99% for all tested cancers. Applying a machine learning algorithm could even help locate the tumor, which would be able to be corroborated by conventional tests, such as mammograms or colonoscopy. “The test in effect is pan-cancer and we continue on this front,” explained Dr. Papadopoulos. The researchers believe that CancerSEEK could cost less than $ 500, similar to current screening cancer tests and that its specificity and sensitivity may be improved by combination with additional tumor biomarkers. “We have initiated prospective studies to test CancerSEEK. This is the way to really evaluate the clinical applicability of the test,” concluded Dr. Papadopoulos when discussing the test’s future prospects.

In 2016 the global liquid biopsy market size was valued at $ 23.49 million, but as money flows in the sector it is predicted to exceed $ 1.2 billion by 2023.

Machine learning is an exponential factor for Next Generation Sequencing


Next-generation sequencing uses machine learning as an exponential factor

In contrast to the detection of a single genetic alteration, NGS aims to record genetic alterations present across diverse cancer subtypes. Moreover, NGS can detect several mutations that could be responsible for driving tumorigenesis and identify resistance mechanisms that may have evolved from pre-existing clones after treatment. As Professor Nicola Normanno, Director of Translational Research Department of the Istituto Nazionale Tumori in Naples, Italy says “ctDNA detection by liquid biopsy is useful, and with the NGS platform, we would be able to obtain a more in-depth molecular characterization of the tumor.” But in a study of NGS liquid biopsy, he and his team highlighted some of the challenges they need to address. “The central point from our study is that we need to work more on specificity and sensitivity to move NGS liquid biopsy into the clinic. This is due to the low level of ctDNA in blood samples but also due to potential artifacts detected by NGS which are from clonal hematopoiesis rather than from tumor-associated DNA. So, we need to understand more about the biology of ctDNA in order to improve the technology.”

On the other hand, a key advantage of NGS liquid biopsies is the detection of mutations in the blood arising from the entire tumor, overcoming the problem of tumor heterogeneity. It can also be used to monitor treatment response and development of resistance. When asked about other potential uses of NGS liquid biopsy, Professor Normanno discussed the possibility of evaluating tumor mutational burden (TMB), which is being considered as a potential biomarker for response to immunotherapy.

Are we close to a turning point?

The size of the liquid biopsy market shows that we are at an early stage of development. The first commercially available tests will need to prove their effectiveness and convince physicians, clinical centers and international organizations for standardization that they can isolate the signal from the noise and create real value for all the healthcare sector.

The intensive use of machine learning algorithms raises the possibilities for more accurate test results connecting the specific cancer type with the patient’s unique gene identity and offering tailor-made treatment options. Additionally, the fact that it is minimally invasive allows tests to be repeated at the minimum psychological cost on people who bear a huge burden due to their condition.

Forbes – Healthcare