Proteomics-based, multivariate random forest method for prediction of protein separation behavior during cation-exchange chromatography

Lung proteomic biomarkers associated with chronic obstructive pulmonary disease

Identifying protein biomarkers for chronic obstructive pulmonary disease (COPD) has been challenging. Most previous studies have used individual proteins or preselected protein panels measured in blood samples. Mass spectrometry proteomic studies of lung tissue have been based on small sample sizes. We used mass spectrometry proteomic approaches to discover protein biomarkers from 150 lung tissue samples representing COPD cases and controls. Top COPD-associated proteins were identified based on multiple linear regression analysis with false discovery rate (FDR) < 0.05. Correlations between pairs of COPD-associated proteins were examined. Machine learning models were also evaluated to identify potential combinations of protein biomarkers related to COPD. We identified 4,407 proteins passing quality controls. Twenty-five proteins were significantly associated with COPD at FDR < 0.05, including interleukin 33, ferritin (light chain and heavy chain), and two proteins related to caveolae (CAV1 and CAVIN1). Multiple previously reported plasma protein biomarkers for COPD were not significantly associated with proteomic analysis of COPD in lung tissue, although RAGE was borderline significant. Eleven pairs of top significant proteins were highly correlated (r > 0.8), including several strongly correlated with RAGE (EHD2 and CAVIN1). Machine learning models using Random Forests with the top 5% of protein biomarkers demonstrated reasonable accuracy (0.707) and area under the curve (0.714) for COPD prediction. Mass spectrometry-based proteomic analysis of lung tissue is a promising approach for the identification of biomarkers for COPD.

A High Throughput Approach to Reconstruct Partial-Body and Neutron Radiation Exposures on an Individual Basis

Biodosimetry-based individualized reconstruction of complex irradiation scenarios (partial-body shielding and/or neutron + photon mixtures) can improve treatment decisions after mass-casualty radiation-related incidents. We used a high-throughput micronucleus assay with automated scanning and imaging software on ex-vivo irradiated human lymphocytes to: a) reconstruct partial-body and/or neutron exposure, and b) estimate separately the photon and neutron doses in a mixed exposure. The mechanistic background is that, compared with total-body photon irradiations, neutrons produce more heavily-damaged lymphocytes with multiple micronuclei/binucleated cell, whereas partial-body exposures produce fewer such lymphocytes. To utilize these differences for biodosimetry, we developed metrics that describe micronuclei distributions in binucleated cells and serve as predictors in machine learning or parametric analyses of the following scenarios: (A) Homogeneous gamma-irradiation, mimicking total-body exposures, vs. mixtures of irradiated blood with unirradiated blood, mimicking partial-body exposures. (B) X rays vs. various neutron + photon mixtures. The results showed high accuracies of scenario and dose reconstructions. Specifically, receiver operating characteristic curve areas (AUC) for sample classification by exposure type reached 0.931 and 0.916 in scenarios A and B, respectively. R² for actual vs. reconstructed doses in these scenarios reached 0.87 and 0.77, respectively. These encouraging findings demonstrate a proof-of-principle for the proposed approach of high-throughput reconstruction of clinically-relevant complex radiation exposure scenarios.

Accounting for host cell protein behavior in anion-exchange chromatography

Host cell proteins (HCP) are a problematic set of impurities in downstream processing (DSP) as they behave most similarly to the target protein during separation. Approaching DSP with the knowledge of HCP separation behavior would be beneficial for the production of high purity recombinant biologics. Therefore, this work was aimed at characterizing the separation behavior of complex mixtures of HCP during a commonly used method: anion-exchange chromatography (AEX). An additional goal was to evaluate the performance of a statistical methodology, based on the characterization data, as a tool for predicting protein separation behavior. Aqueous two-phase partitioning followed by two-dimensional electrophoresis provided data on the three physicochemical properties most commonly exploited during DSP for each HCP: pI (isoelectric point), molecular weight, and surface hydrophobicity. The protein separation behaviors of two alternative expression host extracts (corn germ and E. coli) were characterized. A multivariate random forest (MVRF) statistical methodology was then applied to the database of characterized proteins creating a tool for predicting the AEX behavior of a mixture of proteins. The accuracy of the MVRF method was determined by calculating a root mean squared error value for each database. This measure never exceeded a value of 0.045 (fraction of protein populating each of the multiple separation fractions) for AEX. © 2016 American Institute of Chemical Engineers Biotechnol. Prog., 32:1453-1463, 2016.