CONFERENCE UPDATE: AASLD2020
The potential of liquid biopsy in diagnosing and monitoring NASH patients
Non-alcoholic fatty liver disease (NAFLD) is a common liver disease on an increasing trend associated with the increasing incidence of obesity and type 2 diabetes.1 NAFLD is characterized by hepatic lipid accumulation that often progresses to nonalcoholic steatohepatitis (NASH), liver fibrosis, or cirrhosis.1 While liver biopsy is the gold standard used for NASH diagnosis, the invasive procedure is subjected to risks of surgical complications and may increase the risk of cancer ‘seeding’ to other sites.2 However, there is currently no single noninvasive method that can accurately capture the 4 pathologic components of steatosis, inflammation, hepatocyte ballooning, and fibrosis which are relevant to the mechanisms targeted in the drug development for NASH.
Proteomics, which is the investigation of all proteins expressed by a cell, can be used for protein profiling, alternate splicing and post-translational modifications.1 Using modified-aptamer proteomics, Dr. Rachel Ostroff and her team from SomaLogic, Colorado, United States aimed to design various protein models that could noninvasively diagnose or monitor the progression of NASH. Scanning approximately 5,000 proteins from 2,852 serum samples of 636 participants from the NASH clinical research network natural history cohort, and 2 clinical trials of PIVENS and FLINT, a total of ~15 million protein measurements were obtained. Liver biopsy results were then modelled with the collected protein measurements using machine learning methods. Half of the collected data from each cohort were then used for model training with the other half used for model validation.
After each biopsy component has been independently fed into the machine learning model, 35 unique proteins have been identified and 4 protein models that aligned with liver biopsy findings were created. The 4 protein models were all sufficiently sensitive and precise to characterize the time course and extent of three known drug mechanisms targeted in the treatment of NASH: i) fibrosis: area under the curve (AUC) 0.92/0.85, ii) steatosis: AUC 0.95/0.79, iii) inflammation: AUC 0.83/0.72, and iv) hepatocyte ballooning: AUC 0.87/0.83. By having a concurrent positive score for steatosis, inflammation and ballooning, a biopsy diagnosis of NASH can reach an accuracy of 73%.
In addition to the 4 protein models, the identified metabolic proteins were associated with fatty acid degradation, glucose metabolism, oxidative stress, inflammation, and extracellular matrix degradation. These proteins can reflect the dynamic nature of liver disease and the liver’s ability to heal itself with a high degree of accuracy. Together with the model, these tests may assist new drug development and hold promise for both screening patients and monitoring patient responses to therapy.
That said, these serum protein scanning tests require labor intensive and reliable protein quantification in the blood.1 While NASH may now be supported with these new techniques, to date proteomic studies have provided limited contribution to the drug development for NAFLD.1 In fact, most of the treatment targets for NAFLD were only identified through animal studies with supplementary support from human physiology or epidemiological studies.1
In future, these protein models could be further trained and developed in the general population or in the primary care setting so that adjustments could be made for confounding factors such as age, medications and comorbidities.1 With further data analysis, these protein tests may help medical intervention decisions for both NASH and NAFLD patients.
Translational potential of liquid biopsies for cancer prognostication, profiling and monitoring
The limitations of tissue biopsies include underrepresentation of tumor heterogeneity and difficulty to obtain sufficient material of adequate quality.1 In contrast, liquid biopsies offer a safer, non-invasive alternative to assess tumor specific aberrations through circulating free DNA (cfDNA) that