Transparent Exploration of Machine Learning for Biomarker Discovery from Proteomics and Omics Data

Article Biochemical Research Methods

Interpretation of the DOME Recommendations for Machine Learning in Proteomics and Metabolomics

Magnus Palmblad et al.

Summary: This article discusses the specific guidelines for conducting and reporting machine learning in proteomics and metabolomics, emphasizing the importance of describing molecular structure, function, and physicochemical properties.

JOURNAL OF PROTEOME RESEARCH (2022)

Add to Collection

Article Biochemistry & Molecular Biology

Noninvasive proteomic biomarkers for alcohol-related liver disease

Lili Niu et al.

Summary: This study used mass spectrometry analysis to investigate the liver and plasma proteomics of patients with alcohol-related liver disease. Noninvasive biomarkers associated with early stages of disease progression were identified. Machine learning models revealed that these biomarkers could accurately detect fibrosis and inflammation, and predict future liver-related events and all-cause mortality.

NATURE MEDICINE (2022)

Add to Collection

Review Biochemical Research Methods

How (Not) to Generate a Highly Predictive Biomarker Panel Using Machine Learning

Heather Desaire

Summary: This review warns researchers not to use a biased strategy to select features in order to make their proteomics data look more impressive. The article also cautions against test data leakage into the feature selection step in biomarker studies.

JOURNAL OF PROTEOME RESEARCH (2022)

Add to Collection

Article Cell Biology

Proteome profiling of cerebrospinal fluid reveals biomarker candidates for Parkinson's disease

Ozge Karayel et al.

Summary: The study presents a mass spectrometry-based workflow for analyzing the CSF proteome of Parkinson's disease (PD) patients. The workflow identifies several proteins that are altered in PD patients or significantly correlated with clinical scores, and reveals enhanced neuroinflammation in LRRK2 G2019S carriers. Comparison with urinary proteomes also demonstrates overlapping PD-associated changes, including lysosomal protein alterations, shedding light on the understanding of PD pathogenesis.

CELL REPORTS MEDICINE (2022)

Add to Collection

Article Medicine, Research & Experimental

Urinary proteome profiling for stratifying patients with familial Parkinson's disease

Sebastian Virreira Winter et al.

Summary: The article presents a mass spectrometry-based proteomic workflow for urinary proteome profiling as a valuable strategy for biomarker discovery and patient stratification in Parkinson's disease (PD). The study found significant differences in urinary proteome between PD patients and healthy controls, as well as between LRRK2 G2019S carriers and non-carriers, with lysosomal dysregulation in LRRK2 G2019S mutation carriers. When combined with machine learning, the urinary proteome data alone could effectively classify mutation status and disease manifestation in PD patients.

EMBO MOLECULAR MEDICINE (2021)

Add to Collection

Article Cell Biology

Reproducibility in machine learning for health research: Still a ways to go

Matthew B. A. McDermott et al.

Summary: Research shows that machine learning for health lacks reproducibility compared to other areas, highlighting the need for improvement in this area.

SCIENCE TRANSLATIONAL MEDICINE (2021)

Add to Collection

Article Medicine, Research & Experimental

High-resolution serum proteome trajectories in COVID-19 reveal patient-specific seroconversion

Philipp E. Geyer et al.

Summary: The study utilized proteomics to analyze serum proteomes of COVID-19 patients and controls, finding significant changes in 26% of proteins in COVID-19 patients, including proteins related to the immune system and coagulation mechanisms. The research also revealed correlations between SARS-CoV-2 antibodies and immunoglobulin regions through mass spectrometry-based proteomics, showing potential pathophysiological relevance.

EMBO MOLECULAR MEDICINE (2021)

Add to Collection

Article Biochemistry & Molecular Biology

A time-resolved proteomic and prognostic map of COVID-19

Vadim Demichev et al.

Summary: This study investigates the time-dependent progression of COIVD-19 inpatients by measuring diagnostic parameters and plasma proteomes, revealing an initial spike in inflammatory response followed by tissue repair and immunomodulation. Prognostic marker signatures for treatment strategies were identified and machine learning models based on proteome show transferability to independent cohorts.

CELL SYSTEMS (2021)

Add to Collection

Article Biochemical Research Methods