The purpose of this study was to use machine learning to explore a more precise classification of NAFLD subgroups towards informing individualized therapy.

Clinical characteristics of NAFLD are heterogenous, but current classification for diagnosis is simply based on pathological examination. The conventional pathological classification is insufficient to reflect the complexity and heterogeneity of NAFLD and can not predict the prognosis. Towards precision treatment, a more refined metabolic classification of NAFLD phenotypes is highly demanded for a personalized diagnosis, aiming to identify patients at elevated risk of cardiovascular disease or cirrhosis. This kind of refined classification can provide a more precise diagnosis and enable more individualized preventive interventions and early treatments. In a cross-sectional cohort, unsupervised machine learning was used to cluster patients with biopsy-proved NAFLD from Drum Tower Hospital Affiliated to Nanjing University Medical School based on clinical variables. Verification of the clustering was performed in a longitudinal cohort.

Inclusion Criteria:

• biopsy-proved NALD cohort:

  1. age 18 to 75 years
  2. receiving liver biopsy at the time of metabolic surgery
  3. relatively complete clinical information, including physical examination, biochemical and haematological assessments

• longitudinal cohort

  1. age 18 to 75 years
  2. receiving abdominal imaging examinations,
  3. relatively complete clinical information, including physical examination, biochemical and haematological assessments (4）follow-up time at least more than 12 months

Exclusion Criteria:

• (1)consumed excessive alcohol (≥140 g/week for males or ≥ 70 g/week for females) •
• (2) with history of other liver diseases including chronic hepatitis, biliary obstructive diseases or autoimmune hepatitis

The study enrolled subjects who underwent liver biopsy or physical examination in Drum Tower Hospital affiliated to Nanjing University Medical School.