PREreview del Proteomic profiling of advanced hepatocellular carcinoma identifies predictive signatures of response to treatments

Review of Proteomic profiling of advanced hepatocellular carcinoma identifies predictive signatures of response to treatments

(bioRxiv preprint, DOI: 10.1101/2025.01.03.631224)

Reviewer: Grifton Tafadzwa Muchovu, MSc Student in Medical Biotechnology, University of Piemonte Orientale

Summary

This paper explored how the protein composition of liver cancer tumours (HCC) can serve to predict patients likely to respond to systemic drugs like atezolizumab/bevacizumab or sorafenib. The authors revealed protein signatures associated with treatment response in diagnostic biopsy specimens using high-resolution mass spectrometry and emphasized the possibility of an important role of mitochondrial oxidative phosphorylation and immune infiltration in the resistance.

Key Strengths

• Clinical significance: The ability to predict the response to treatment is one of the key unmet clinical needs in the advanced HCC. The demonstration of the ability of diagnostic biopsies to be employed in proteomic profiling is a good start on the road to practical implementation.

• Technical method: Deep proteomics and machine-learning can be used effectively to discover biomarkers, and more importantly, they can be used to show how advanced analytics can find significant patterns in a small amount of clinical data.

• Mechanistic understanding: The relationship between metabolic reprogramming and immune exclusion is interesting and offers a possible biological rationale of treatment refusal.

Points to Improve or Clarify

• Cohort size and validation: The sample of patients (e.g. 9 and 15 non-responders in atezolizumab/bevacizumab group) is small. Such protein signatures would be useful in confirming whether they have a predictive value when pertaining to an independent and larger cohort that would be external validation.

• Etiopathogenesis of liver disease: Since HCC has many etiologies (e.g. metabolic dysfunction like MASH, alcohol-related liver disease), it would be appropriate to examine whether the proteomic signatures are etiology-specific or, at least, comment on this as a possible source of variability.

• Machine-learning description: The choice of model, feature ranking, and cross-validation information would be better disclosed and the results could be easier reproducible and evaluated (or access to code).

• Follow-up: The experiments with the 3D spheroid are a good beginning and further data on in-vivo experiments would help to make the point that oxidative phosphorylation has a direct effect on immune infiltration and treatment response.

Minor Comments

• Elaborate in the discussion on the issue of whether proteins that are common in treatment groups are indicative of general HCC biology or common resistance pathways.

Overall Assessment

It is a good and clinically applicable proteomic research. It generates promising biomarker candidates and mechanistic hypotheses which may eventually inform the selection of therapy in advanced HCC. This will be followed by more important steps of larger, independent validation and more open machine-learning reporting before clinical implementation.

Competing interests

The author declares that they have no competing interests.