Predicting immune checkpoint therapy response in three independent metastatic melanoma cohorts.

Introduction: While Immune checkpoint inhibition (ICI) therapy shows significant efficacy in metastatic melanoma, only about 50% respond, lacking reliable predictive methods. We introduce a panel of six proteins aimed at predicting response to ICI therapy. Methods: Evaluating previously reported proteins in two untreated melanoma cohorts, we used a published predictive model (EaSIeR score) to identify potential proteins distinguishing responders and non-responders. Results: Six proteins initially identified in the ICI cohort correlated with predicted response in the untreated cohort. Additionally, three proteins correlated with patient survival, both at the protein, and at the transcript levels, in an independent immunotherapy treated cohort. Discussion: Our study identifies predictive biomarkers across three melanoma cohorts, suggesting their use in therapeutic decision-making.

Predicting immune checkpoint therapy response in three independent metastatic melanoma cohorts.

While Immune checkpoint inhibition (ICI) therapy shows significant efficacy in metastatic melanoma, only about 50% respond, lacking reliable predictive methods. We introduce a panel of six proteins aimed at predicting response to ICI therapy. Evaluating previously reported proteins in two untreated melanoma cohorts, we used a published predictive model (EaSIeR score) to identify potential proteins distinguishing responders and non-responders. Six proteins initially identified in the ICI cohort correlated with predicted response in the untreated cohort. Additionally, three proteins correlated with patient survival, both at the protein, and at the transcript levels, in an independent immunotherapy treated cohort. Our study identifies predictive biomarkers across three melanoma cohorts, suggesting their use in therapeutic decision-making.

Identifying Ferroptosis Inducers, HDAC, and RTK Inhibitor Sensitivity in Melanoma Subtypes through Unbiased Drug Target Prediction.

The utilization of PD1 and CTLA4 inhibitors has revolutionized the treatment of malignant melanoma (MM). However, resistance to targeted and immune-checkpoint-based therapies still poses a significant problem. Here we mine large scale MM proteogenomic data integrating it with MM cell line dependency screen, and drug sensitivity data to identify druggable targets and forecast treatment efficacy and resistance. Leveraging protein profiles from established MM subtypes and molecular structures of 82 cancer treatment drugs, we identified nine candidate hub proteins, mTOR, FYN, PIK3CB, EGFR, MAPK3, MAP4K1, MAP2K1, SRC and AKT1, across five distinct MM subtypes. These proteins serve as potential drug targets applicable to one or multiple MM subtypes. By analyzing transcriptomic data from 48 publicly accessible melanoma cell lines sourced from Achilles and CRISPR dependency screens, we forecasted 162 potentially targetable genes. We also identified genetic resistance in 260 genes across at least one melanoma subtype. In addition, we employed publicly available compound sensitivity data (Cancer Therapeutics Response Portal, CTRPv2) on the cell lines to assess the correlation of compound effectiveness within each subtype. We have identified 20 compounds exhibiting potential drug impact in at least one melanoma subtype. Remarkably, employing this unbiased approach, we have uncovered compounds targeting ferroptosis, that demonstrate a striking 30x fold difference in sensitivity among different subtypes. This implies that the proteogenomic classification of melanoma has the potential to predict sensitivity to ferroptosis compounds. Our results suggest innovative and novel therapeutic strategies by stratifying melanoma samples through proteomic profiling, offering a spectrum of novel therapeutic interventions and prospects for combination therapy. Highlights: (1) Proteogenomic subtype classification can define the landscape of genetic dependencies in melanoma (2) Nine proteins from molecular subtypes were identified as potential drug targets for specified MM patients (3) 20 compounds identified that show potential effectiveness in at least one melanoma subtype (4) Proteogenomics can predict specific ferroptosis inducers, HDAC, and RTK Inhibitor sensitivity in melanoma subtypes.

WOMBAT-P: Benchmarking Label-Free Proteomics Data Analysis Workflows.

The inherent diversity of approaches in proteomics research has led to a wide range of software solutions for data analysis. These software solutions encompass multiple tools, each employing different algorithms for various tasks such as peptide-spectrum matching, protein inference, quantification, statistical analysis, and visualization. To enable an unbiased comparison of commonly used bottom-up label-free proteomics workflows, we introduce WOMBAT-P, a versatile platform designed for automated benchmarking and comparison. WOMBAT-P simplifies the processing of public data by utilizing the sample and data relationship format for proteomics (SDRF-Proteomics) as input. This feature streamlines the analysis of annotated local or public ProteomeXchange data sets, promoting efficient comparisons among diverse outputs. Through an evaluation using experimental ground truth data and a realistic biological data set, we uncover significant disparities and a limited overlap in the quantified proteins. WOMBAT-P not only enables rapid execution and seamless comparison of workflows but also provides valuable insights into the capabilities of different software solutions. These benchmarking metrics are a valuable resource for researchers in selecting the most suitable workflow for their specific data sets. The modular architecture of WOMBAT-P promotes extensibility and customization. The software is available at https://github.com/wombat-p/WOMBAT-Pipelines.

Ten quick tips for building FAIR workflows.

Research data is accumulating rapidly and with it the challenge of fully reproducible science. As a consequence, implementation of high-quality management of scientific data has become a global priority. The FAIR (Findable, Accesible, Interoperable and Reusable) principles provide practical guidelines for maximizing the value of research data; however, processing data using workflows-systematic executions of a series of computational tools-is equally important for good data management. The FAIR principles have recently been adapted to Research Software (FAIR4RS Principles) to promote the reproducibility and reusability of any type of research software. Here, we propose a set of 10 quick tips, drafted by experienced workflow developers that will help researchers to apply FAIR4RS principles to workflows. The tips have been arranged according to the FAIR acronym, clarifying the purpose of each tip with respect to the FAIR4RS principles. Altogether, these tips can be seen as practical guidelines for workflow developers who aim to contribute to more reproducible and sustainable computational science, aiming to positively impact the open science and FAIR community.

Age-associated differences in the human lung extracellular matrix.

Extracellular matrix (ECM) remodeling has been associated with chronic lung diseases. However, information about specific age-associated differences in lung ECM is currently limited. In this study, we aimed to identify and localize age-associated ECM differences in human lungs using comprehensive transcriptomic, proteomic, and immunohistochemical analyses. Our previously identified age-associated gene expression signature of the lung was re-analyzed limiting it to an aging signature based on 270 control patients (37-80 years) and focused on the Matrisome core geneset using geneset enrichment analysis. To validate the age-associated transcriptomic differences on protein level, we compared the age-associated ECM genes (false discovery rate, FDR < 0.05) with a profile of age-associated proteins identified from a lung tissue proteomics dataset from nine control patients (49-76 years) (FDR < 0.05). Extensive immunohistochemical analysis was used to localize and semi-quantify the age-associated ECM differences in lung tissues from 62 control patients (18-82 years). Comparative analysis of transcriptomic and proteomic data identified seven ECM proteins with higher expression with age at both gene and protein levels: COL1A1, COL6A1, COL6A2, COL14A1, FBLN2, LTBP4, and LUM. With immunohistochemistry, we demonstrated higher protein levels with age for COL6A2 in whole tissue, parenchyma, airway wall, and blood vessel, for COL14A1 and LUM in bronchial epithelium, and COL1A1 in lung parenchyma. Our study revealed that higher age is associated with lung ECM remodeling, with specific differences occurring in defined regions within the lung. These differences may affect lung structure and physiology with aging and as such may increase susceptibility to developing chronic lung diseases.NEW & NOTEWORTHY We identified seven age-associated extracellular matrix (ECM) proteins, i.e., COL1A1, COL6A1, COL6A2 COL14A1, FBLN2, LTBP4, and LUM with higher transcript and protein levels in human lung tissue with age. Extensive immunohistochemical analysis revealed significant age-associated differences for COL6A2 in whole tissue, parenchyma, airway wall, and vessel, for COL14A1 and LUM in bronchial epithelium, and COL1A1 in parenchyma. Our findings lay a new foundation for the investigation of ECM differences in age-associated chronic lung diseases.

A proteomics approach to identify COPD-related changes in lung fibroblasts.

Lung fibroblasts are implicated in abnormal tissue repair in chronic obstructive pulmonary disease (COPD). The exact mechanisms are unknown and comprehensive analysis comparing COPD- and control fibroblasts is lacking. The aim of this study is to gain insight into the role of lung fibroblasts in COPD pathology using unbiased proteomic and transcriptomic analysis. Protein and RNA were isolated from cultured parenchymal lung fibroblasts of 17 patients with stage IV COPD and 16 non-COPD controls. Proteins were analyzed using LC-MS/MS and RNA through RNA sequencing. Differential protein and gene expression in COPD was assessed via linear regression, followed by pathway enrichment, correlation analysis, and immunohistological staining in lung tissue. Proteomic and transcriptomic data were compared to investigate the overlap and correlation between both levels of data. We identified 40 differentially expressed (DE) proteins and zero DE genes between COPD and control fibroblasts. The most significant DE proteins were HNRNPA2B1 and FHL1. Thirteen of the 40 proteins were previously associated with COPD, including FHL1 and GSTP1. Six of the 40 proteins were related to telomere maintenance pathways, and were positively correlated with the senescence marker LMNB1. No significant correlation between gene and protein expression was observed for the 40 proteins. We hereby describe 40 DE proteins in COPD fibroblasts including previously described COPD proteins (FHL1, GSTP1) and new COPD research targets like HNRNPA2B1. Lack of overlap and correlation between gene and protein data supports the use of unbiased proteomics analysis and indicates that different types of information are generated with both methods.

Biotransformation of Trastuzumab and Pertuzumab in Breast Cancer Patients Assessed by Affinity Enrichment and Ion-Exchange Chromatography.

Therapeutic proteins (TPs) are known to be heterogeneous due to modifications that occur during the production process and storage. Modifications may also occur in TPs after their administration to patients due to in vivo biotransformation. Ligand binding assays, which are widely used in the bioanalysis of TPs in body fluids, are typically unable to distinguish such modifications. Liquid chromatography coupled to mass spectrometry is being increasingly used to study modifications in TPs, but its use to study in vivo biotransformation has been limited until now. We present a novel approach that combines affinity enrichment using Affimer reagents with ion-exchange chromatography (IEX) to analyze charge variants of the TPs trastuzumab and pertuzumab in plasma of patients undergoing therapy for HER2-positive breast cancer. Affimer reagents were immobilized via engineered Cys tags to maleimide beads, and the TPs were eluted under acidic conditions followed by rapid neutralization. The enriched TPs were analyzed by cation-exchange chromatography (IEX) using pH-gradient elution, resulting in the separation of about 20 charge variants for trastuzumab and about five charge variants for pertuzumab. A comparison between in vitro stressed TPs spiked into plasma, and TPs enriched from patient plasma showed that the observed profiles were highly similar. This indicates that in vitro stress testing in plasma can mimic the situation in patient plasma, as far as the generation of charge variants is concerned. SIGNIFICANCE STATEMENT: This research attempts to elucidate the modifications that occur in therapeutic proteins (TPs) after they have been administered to patients. This is important because there is little knowledge about the fate of TPs in this regard, and certain modifications could affect their efficiency. Our results show that the modifications discovered are most likely due to a chemical process and are not patient specific.

GeneNetTools: tests for Gaussian graphical models with shrinkage.

MOTIVATION: Gaussian graphical models (GGMs) are network representations of random variables (as nodes) and their partial correlations (as edges). GGMs overcome the challenges of high-dimensional data analysis by using shrinkage methodologies. Therefore, they have become useful to reconstruct gene regulatory networks from gene-expression profiles. However, it is often ignored that the partial correlations are 'shrunk' and that they cannot be compared/assessed directly. Therefore, accurate (differential) network analyses need to account for the number of variables, the sample size, and also the shrinkage value, otherwise, the analysis and its biological interpretation would turn biased. To date, there are no appropriate methods to account for these factors and address these issues. RESULTS: We derive the statistical properties of the partial correlation obtained with the Ledoit-Wolf shrinkage. Our result provides a toolbox for (differential) network analyses as (i) confidence intervals, (ii) a test for zero partial correlation (null-effects) and (iii) a test to compare partial correlations. Our novel (parametric) methods account for the number of variables, the sample size and the shrinkage values. Additionally, they are computationally fast, simple to implement and require only basic statistical knowledge. Our simulations show that the novel tests perform better than DiffNetFDR-a recently published alternative-in terms of the trade-off between true and false positives. The methods are demonstrated on synthetic data and two gene-expression datasets from Escherichia coli and Mus musculus. AVAILABILITY AND IMPLEMENTATION: The R package with the methods and the R script with the analysis are available in https://github.com/V-Bernal/GeneNetTools. SUPPLEMENTARY INFORMATION: Supplementary data are available at Bioinformatics online.

In-depth proteomic analysis reveals unique subtype-specific signatures in human small-cell lung cancer.

BACKGROUND: Small-cell lung cancer (SCLC) molecular subtypes have been primarily characterized based on the expression pattern of the following key transcription regulators: ASCL1 (SCLC-A), NEUROD1 (SCLC-N), POU2F3 (SCLC-P) and YAP1 (SCLC-Y). Here, we investigated the proteomic landscape of these molecular subsets with the aim to identify novel subtype-specific proteins of diagnostic and therapeutic relevance. METHODS: Pellets and cell media of 26 human SCLC cell lines were subjected to label-free shotgun proteomics for large-scale protein identification and quantitation, followed by in-depth bioinformatic analyses. Proteomic data were correlated with the cell lines' phenotypic characteristics and with public transcriptomic data of SCLC cell lines and tissues. RESULTS: Our quantitative proteomic data highlighted that four molecular subtypes are clearly distinguishable at the protein level. The cell lines exhibited diverse neuroendocrine and epithelial-mesenchymal characteristics that varied by subtype. A total of 367 proteins were identified in the cell pellet and 34 in the culture media that showed significant up- or downregulation in one subtype, including known druggable proteins and potential blood-based markers. Pathway enrichment analysis and parallel investigation of transcriptomics from SCLC cell lines outlined unique signatures for each subtype, such as upregulated oxidative phosphorylation in SCLC-A, DNA replication in SCLC-N, neurotrophin signalling in SCLC-P and epithelial-mesenchymal transition in SCLC-Y. Importantly, we identified the YAP1-driven subtype as the most distinct SCLC subgroup. Using sparse partial least squares discriminant analysis, we identified proteins that clearly distinguish four SCLC subtypes based on their expression pattern, including potential diagnostic markers for SCLC-Y (e.g. GPX8, PKD2 and UFO). CONCLUSIONS: We report for the first time, the protein expression differences among SCLC subtypes. By shedding light on potential subtype-specific therapeutic vulnerabilities and diagnostic biomarkers, our results may contribute to a better understanding of SCLC biology and the development of novel therapies.

Comparative Assessment of Quantification Methods for Tumor Tissue Phosphoproteomics.

With increasing sensitivity and accuracy in mass spectrometry, the tumor phosphoproteome is getting into reach. However, the selection of quantitation techniques best-suited to the biomedical question and diagnostic requirements remains a trial and error decision as no study has directly compared their performance for tumor tissue phosphoproteomics. We compared label-free quantification (LFQ), spike-in-SILAC (stable isotope labeling by amino acids in cell culture), and tandem mass tag (TMT) isobaric tandem mass tags technology for quantitative phosphosite profiling in tumor tissue. Compared to the classic SILAC method, spike-in-SILAC is not limited to cell culture analysis, making it suitable for quantitative analysis of tumor tissue samples. TMT offered the lowest accuracy and the highest precision and robustness toward different phosphosite abundances and matrices. Spike-in-SILAC offered the best compromise between these features but suffered from a low phosphosite coverage. LFQ offered the lowest precision but the highest number of identifications. Both spike-in-SILAC and LFQ presented susceptibility to matrix effects. Match between run (MBR)-based analysis enhanced the phosphosite coverage across technical replicates in LFQ and spike-in-SILAC but further reduced the precision and robustness of quantification. The choice of quantitative methodology is critical for both study design such as sample size in sample groups and quantified phosphosites and comparison of published cancer phosphoproteomes. Using ovarian cancer tissue as an example, our study builds a resource for the design and analysis of quantitative phosphoproteomic studies in cancer research and diagnostics.

Age-specific oncogenic pathways in head and neck squamous cell carcinoma - are elderly a different subcategory?

BACKGROUND: In recent clinical practice, an increasing number of elderly patients suffering from head and neck squamous cell carcinoma (HNSCC) of unknown pathophysiology is observed. The majority of HNSCC patients can roughly be divided into three subcategories. First, a small group of young patients who present with variants of genomic aberrations and inheritable diseases like Fanconi anaemia. Second, an increasing population of HPV-related HNSCCs that are regarded as genomic stable tumours with a more favourable prognosis. Though HPV-related tumours used to be more common among younger males, a notable rise in the elderly population is observed. The third subcategory, that of HPV-negative tumours, has been shown to be more heterogeneous with involvement of a variety of oncogenic pathways related to lifestyle factors like smoking and alcohol consumption, often seen in middle-aged males. Some of these pathways could be related to age, such as TP53 alterations, EGFR activation, apoptotic pathway alterations and field cancerization. CONCLUSIONS: In this narrative review, we provide an overview of established and newly discovered age-specific pathophysiological mechanisms underlying HNSCC. We propose a fourth subcategory of patients with a suspected different pathophysiology: elderly (HPV-negative) HNSCC patients without a history of tobacco and alcohol consumption. In this subcategory, carcinogenesis seems to be a multi-step process based on genomic instability, immunosenescence, cell cycle disruption and telomere shortening. To conclude, we discuss suggestions for future research to fill the knowledge gap about age-dependent HNSCC carcinogenesis.

Plasma sRAGE levels strongly associate with centrilobular emphysema assessed by HRCT scans.

BACKGROUND: There is a strong need for biomarkers to better characterize individuals with COPD and to take into account the heterogeneity of COPD. The blood protein sRAGE has been put forward as promising biomarker for COPD in general and emphysema in particular. Here, we measured plasma sRAGE levels using quantitative LC-MS and assessed whether the plasma sRAGE levels associate with (changes in) lung function, radiological emphysema parameters, and radiological subtypes of emphysema. METHODS: Three hundred and twenty-four COPD patients (mean FEV1: 63%predicted) and 185 healthy controls from the COPDGene study were selected. Plasma sRAGE was measured by immunoprecipitation in 96-well plate methodology to enrich sRAGE, followed by targeted quantitative liquid chromatography-mass spectrometry. Spirometry and HRCT scans (inspiration and expiration) with a 5-year follow-up were used; both subjected to high quality control standards. RESULTS: Lower sRAGE values significantly associated with the presence of COPD, the severity of airflow obstruction, the severity of emphysema on HRCT, the heterogeneous distribution of emphysema, centrilobular emphysema, and 5-year progression of emphysema. However, sRAGE values did not associate with airway wall thickness or paraseptal emphysema. CONCLUSIONS: Rather than being a general COPD biomarker, sRAGE is especially a promising biomarker for centrilobular emphysema. Follow-up studies should elucidate whether sRAGE can be used as a biomarker for other COPD phenotypes as well.

Accurate Prediction of Protein Sequences for Proteogenomics Data Integration.

This book chapter discusses proteogenomics data integration and provides an overview into the different omics layer involved in defining the proteome of a living organism. Various aspects of genome variability affecting either the sequence or abundance level of proteins are discussed in this book chapter, such as the effect of single-nucleotide variants or larger genomic structural variants on the proteome. Next, various sequencing technologies are introduced and discussed from a proteogenomics data integration perspective such as those providing short- and long-read sequencing and listing their respective advantages and shortcomings for accurate protein variant prediction using genomic/transcriptomics sequencing data. Finally, the various bioinformatics tools used to process and analyze DNA/RNA sequencing data are discussed with the ultimate goal of obtaining accurately predicted sample-specific protein sequences that can be used as a drop-in replacement in existing approaches for peptide and protein identification using popular database search engines such as MSFragger, SearchGUI/PeptideShaker.

Deep Proteomic Analysis on Biobanked Paraffine-Archived Melanoma with Prognostic/Predictive Biomarker Read-Out.

The discovery of novel protein biomarkers in melanoma is crucial. Our introduction of formalin-fixed paraffin-embedded (FFPE) tumor protocol provides new opportunities to understand the progression of melanoma and open the possibility to screen thousands of FFPE samples deposited in tumor biobanks and available at hospital pathology departments. In our retrospective biobank pilot study, 90 FFPE samples from 77 patients were processed. Protein quantitation was performed by high-resolution mass spectrometry and validated by histopathologic analysis. The global protein expression formed six sample clusters. Proteins such as TRAF6 and ARMC10 were upregulated in clusters with enrichment for shorter survival, and proteins such as AIFI1 were upregulated in clusters with enrichment for longer survival. The cohort's heterogeneity was addressed by comparing primary and metastasis samples, as well comparing clinical stages. Within immunotherapy and targeted therapy subgroups, the upregulation of the VEGFA-VEGFR2 pathway, RNA splicing, increased activity of immune cells, extracellular matrix, and metabolic pathways were positively associated with patient outcome. To summarize, we were able to (i) link global protein expression profiles to survival, and they proved to be an independent prognostic indicator, as well as (ii) identify proteins that are potential predictors of a patient's response to immunotherapy and targeted therapy, suggesting new opportunities for precision medicine developments.

Olympic anti-doping laboratory: the analytical technological road from 2016 Rio De Janeiro to 2021 Tokyo.

The summer Olympic Games is the major mega sports event since the first modern era Olympiad, held in Athens, Greece in 1896. International Olympic Committee (IOC) has the responsibility of the organization of the summer and winter Games ensuring the broadcast in all corners of earth. The World Anti-Doping Agency (WADA) is the responsible organization of the fight against doping in sports. IOC and WADA support the event's country WADA Accredited Laboratory to incorporate the maximum of the new analytical technologies to become applicable during the event's antidoping testing. The current study reviewed the last 5 years progresses of the antidoping system with emphasis on the laboratory field.

Adsorptive Microtiter Plates As Solid Supports in Affinity Purification Workflows.

Affinity ligands such as antibodies are widely used in (bio)medical research for purifying proteins from complex biological samples. These ligands are generally immobilized onto solid supports which facilitate the separation of a captured protein from the sample matrix. Adsorptive microtiter plates are commonly used as solid supports prior to immunochemical detection (e.g., immunoassays) but hardly ever prior to liquid chromatography-mass spectrometry (LC-MS-)-based detection. Here, we describe the use of adsorptive microtiter plates for protein enrichment prior to LC-MS detection, and we discuss opportunities and challenges of corresponding workflows, based on examples of targeted (i.e., soluble receptor for advanced glycation end-products (sRAGE) in human serum) and discovery-based workflows (i.e., transcription factor p65 (NF-κB) in lysed murine RAW 264.7 macrophages and peptidyl-prolyl cis-trans isomerase FKBP5 (FKBP5) in lysed human A549 alveolar basal epithelial cells). Thereby, we aim to highlight the potential usefulness of adsorptive microtiter plates in affinity purification workflows prior to LC-MS detection, which could increase their usage in mass spectrometry-based protein research.

Combined Metabolic and Chemical (CoMetChem) Labeling Using Stable Isotopes-a Strategy to Reveal Site-Specific Histone Acetylation and Deacetylation Rates by LC-MS.

Histone acetylation is an important, reversible post-translational protein modification and a hallmark of epigenetic regulation. However, little is known about the dynamics of this process, due to the lack of analytical methods that can capture site-specific acetylation and deacetylation reactions. We present a new approach that combines metabolic and chemical labeling (CoMetChem) using uniformly 13C-labeled glucose and stable isotope-labeled acetic anhydride. Thereby, chemically equivalent, fully acetylated histone species are generated, enabling accurate relative quantification of site-specific lysine acetylation dynamics in tryptic peptides using high-resolution mass spectrometry. We show that CoMetChem enables site-specific quantification of the incorporation or loss of lysine acetylation over time, allowing the determination of reaction rates for acetylation and deacetylation. Thus, the CoMetChem methodology provides a comprehensive description of site-specific acetylation dynamics.

The 'un-shrunk' partial correlation in Gaussian graphical models.

BACKGROUND: In systems biology, it is important to reconstruct regulatory networks from quantitative molecular profiles. Gaussian graphical models (GGMs) are one of the most popular methods to this end. A GGM consists of nodes (representing the transcripts, metabolites or proteins) inter-connected by edges (reflecting their partial correlations). Learning the edges from quantitative molecular profiles is statistically challenging, as there are usually fewer samples than nodes ('high dimensional problem'). Shrinkage methods address this issue by learning a regularized GGM. However, it remains open to study how the shrinkage affects the final result and its interpretation. RESULTS: We show that the shrinkage biases the partial correlation in a non-linear way. This bias does not only change the magnitudes of the partial correlations but also affects their order. Furthermore, it makes networks obtained from different experiments incomparable and hinders their biological interpretation. We propose a method, referred to as 'un-shrinking' the partial correlation, which corrects for this non-linear bias. Unlike traditional methods, which use a fixed shrinkage value, the new approach provides partial correlations that are closer to the actual (population) values and that are easier to interpret. This is demonstrated on two gene expression datasets from Escherichia coli and Mus musculus. CONCLUSIONS: GGMs are popular undirected graphical models based on partial correlations. The application of GGMs to reconstruct regulatory networks is commonly performed using shrinkage to overcome the 'high-dimensional problem'. Besides it advantages, we have identified that the shrinkage introduces a non-linear bias in the partial correlations. Ignoring this type of effects caused by the shrinkage can obscure the interpretation of the network, and impede the validation of earlier reported results.