Machine Learning‐Assisted SERS Quantification of Sialylated Alpha‐Fetoprotein: From Single‐Cell Analysis to Hepatocellular Carcinoma Risk Assessment

Sialylated alpha-fetoprotein (sAFP) is a very potential marker for the pathogenesis exploration and clinical assessment of hepatocellular carcinoma (HCC). The specific and sensitive quantification of sAFP across multiple aspects is a primary premise. This work constructs a functionalized gold/silver nanocube-encapsulated microgel (Au/AgNC-MG), which can specifically capture sAFP through dual aptamer-based recognition and generate sensitive Raman fingerprints through the heterogeneous bimetallic SERS system. To further improve the specificity and quantifiability of sAFP detection in different scenarios, a series of machine learning (ML) algorithms, including a sAFP classification algorithm, a sAFP image-processing algorithm, and a sAFP-based clinical HCC risk assessment algorithm, were established for sAFP quantification, imaging of single-cell secreted sAFP, and clinical HCC risk assessment from general check-up to cirrhosis clinic patient populations. An online HCC Risk Assessment website is built for the convenience of practical clinical application. The constructed Au/AgNC-MG and established ML algorithms compose a general paradigm for HCC-related laboratory research and clinical applications.