Publication highlights September 2025

Empowering progress in protein biomarker research – a new milestone reached

Olink’s Proximity Extension Assay (PEA) technology is established at the heart of proteomic and multiomic research studies and has now been cited in over 3000 peer-reviewed articles across the broad span of biomedical research.

This is testament to the widespread adoption of PEA by the scientific community, which continues to accelerate, empowering researchers around the World to deliver breakthrough findings of the highest quality and significance. Here we would like to highlight just four recently published articles to exemplify the diversity and outstanding quality represented by this body of work.

If you would like to know more about the latest, on-going research using Olink technology, sign-up for our virtual event, Olink Proteomics World 2025.

Genetic and molecular landscape of comorbidities in people living with HIV

A joint study by between the Hannover Medical School and the Helmholtz Centre for Infection Research used the Olink® Explore 3072 high-throughput proteomics platform in a prospective multiomic study to better understand why people living with HIV (PLHIV) have an increased risk of developing multiple, non-HIV comorbidities.

They applied proteomics, genomics, epigenomics, transcriptomics, proteomics, metabolomics and immune response analysis to identify latent factors (LF – groups of interrelated molecular features) associated with multiple comorbidities.

Key highlights:

• 21 multiomic LFs were identified, accounting for between 1-20% of total variance, and notably, most of these included molecules from all 5 omics sources.
• 19 of 21 LFs were significantly correlated with systemic inflammation.
• LF6 was mostly protein featured (capturing differences in innate immune activation and NF-κB activation) and was associated to presence of carotid plaques.
• Proteogenomic analysis indicated >3,000 pQTLs for >1,000 proteins. When compared with the independent UKB-PPP coort, there was concordance of direction in >99% of pQTLs, indicating a common regulation of protein abundances between PLHIV and healthy people.
• This data has been made available as an open-access resource to the scientific community.

A protein risk score for dementia in individuals with major depressive disorder

The mechanisms linking major depressive disorder (MDD) to an increased risk of Alzheimer’s disease and related dementia (ADRD) are not fully understood. Scientists from the University of Connecticut used data from the UK Biobank Pharma Proteomics Project generated using Olink® Explore 3072 to derive a predictive risk score for incident ADRD in individuals who had a history of MDD. Furthermore, a proteogenomic analysis was also conducted to identify proteins with likely causality in disease development.

Key highlights:

• While almost 500 proteins were significantly associated with ADRD risk in the no-MDD group, just 6 were significant in individuals with a history of MDD – NfL, GFAP, PSG1, VGF, GET3 and HPGDS. GET3 was the only marker unique for MDD patients.
• These associations remained statistically significant after excluding APOEe4 carrier status (the primary genetic risk factor for ADRD).
• Machine learning (LASSO) was applied to develop a protein score (“PrRSMDD-ADRD”) that predicted 10-year risk of ADRD in MDD individuals with a C-statistic of 0.84.
• Mendelian Randomization using proteomic and gene variant data from UKB provided good evidence that both ApoE (protective factor) and IL-10RB (risk factor) are causally linked to ADRD, with hazard ratios of 1.81 & 1.41 per standard deviation change, respectively.

Multiomic models predict treatment outcome in combinational neoadjuvant therapy of breast cancer.

Scientists from Sichuan University, Chengdu, combined RNAseq of tumor biopsy tissue with serum proteomics using the Olink® Target 96 Immuno-oncology panel to gain important insights into therapeutic responses in a clinical trial for combinational neoadjuvant therapy of patients with triple-negative breast cancer (TNBC). Current chemotherapy approaches for TNBC are insufficient, and here they used a combination of chemotherapy, anti-PD1 immunotherapy and anti-angiogenic therapy with a VEGFR2 inhibitor.

The triple-therapy showed a tolerable safety profile and good efficacy in the majority of patients, with a pathological complete response (pCR) rate of ~68%. Tissue RNA and serum protein expression were then compared pre- and post-therapy to identify markers that could predict treatment response at baseline, and/or be used to monitor efficacy during therapy.

Key highlights:

• 31 proteins showed significant changes after treatment, indicating a complex systemic immune response, with cytokine-receptor signaling & chemokine signaling among the most enriched pathways.
• Baseline levels of serum IL-18 and biopsy tissue PD-L1 mRNA could be combined as a model that could predict responders from non-responders with AUC=0.8232. 88% of patients with a high vs low PRP score achieved pCR, making this a potentially valuable stratification tool for therapeutic guidance.
• Looking at changes in protein levels before and after therapy, they identified a 5-protein model to discriminate positive-therapy improvement with an AUC=0.93. This score could be a useful tool to identify patients who may not benefit from continued therapy and to develop potential strategies for increasing the pCR rate.

A candidate protein signature to predict ALS

Researchers from the National Institute on Aging, NIH carried out a study to identify important new biomarkers for  amyotrophic lateral sclerosis (ALS). They used Olink® Explore 3072 to measure plasma proteomics of ALS patients predominantly  in the mostly asymptomatic prodromal phase of the disease, comparing data to a control group that included both healthy subjects and those with non-ALS neurological conditions.  In addition to identifying new candidate biomarkers, the study also shed new light on the prodromal phase of the disease.

Key highlights:

• 33 ALS-specific protein changes were identified  – these included the well-characterized markers, NfL and LIF, but the majority of proteins have not  previously been definitively linked to ALS.
• These markers were replicated in a 2nd cohort, with a remarkable overall concordance of R=0.83.
• Machine learning derived a model comprising 17 proteins & 3 clinical variables that discriminated ALS vs healthy controls with AUC=0.962 and ALS vs neurological + healthy controls with AUC=0.893.
• The model could identify ALS vs controls in the independent UKB-PPP cohort with 99% accuracy, despite the lack of any bridging samples.
• The dynamics of the the ALS risk score suggested that disease process may begin to occur further in advance of ALS diagnosis compared to current thinking.