Michal Marczyk

Adaptive filtering of microarray gene expression data based on Gaussian mixture decomposition

BackgroundDNA microarrays are used for discovery of genes expressed differentially between various biological conditions. In microarray experiments the number of analyzed samples is often much lower than the number of genes (probe sets) which leads to many false discoveries. Multiple testing correction methods control the number of false discoveries but decrease the sensitivity of discovering differentially expressed genes. Concerning this problem, filtering methods for improving the power of detection of differentially expressed genes were proposed in earlier papers. These techniques are two-step procedures, where in the first step some pool of non-informative genes is removed and in the second step only the pool of the retained genes is used for searching for differentially expressed genes.ResultsA very important parameter to choose is the proportion between the sizes of the pools of removed and retained genes. A new method, which we propose, allow to determine close to optimal threshold values for sample means and sample variances for gene filtering. The method is adaptive and based on the decomposition of the histogram of gene expression means or variances into mixture of Gaussian components.ConclusionsBy performing analyses of several publicly available datasets and simulated datasets we demonstrate that our adaptive method increases sensitivity of finding differentially expressed genes compared to previous methods of filtering microarray data based on using fixed threshold values.

Comparison of peptide cancer signatures identified by mass spectrometry in serum of patients with head and neck, lung and colorectal cancers: Association with tumor progression

Mass spectrometry-based analyses of the low-molecular-weight fraction of serum proteome allow identifying proteome profiles (signatures) that are potentially useful in detection and diagnostics of cancer. Here we compared serum proteome profiles of healthy donors and patients with three different types of cancer aiming to identify peptide signatures that were either common for all cancer samples or specific for cancer type. Blood samples were collected before start of the therapy from patients with head and neck squamous cell cancer, colorectal adenocarcinoma and non-small cell lung cancer, and from a corresponding group of healthy volunteers. Mass profiles of the serum proteome were recorded in the range between 2 and 13 kDa using MALDI-ToF spectrometry and 131 identified peptide ions were used for statistical analyses. Similar degrees of overall similarities were observed in all intra-group and inter-group analyses when general features of serum proteome profiles were compared between individual samples. However, classifiers built of selected spectral components allowed differentiation between healthy donors and three groups of cancer patients with 69-74% sensitivity and 82-84% specificity. There were two common peptide species (3766 and 5867 Da) with increased levels in all cancer samples. Several spectral components permitted differentiation between lung cancer samples and either head and neck cancer or colorectal cancer samples, but two latter types of samples could not be properly discriminated. Abundance of spectral components that putatively corresponded to fragments of serum amyloid A (11511 and 11667 Da) was highest in lung cancer samples, yet increased levels of these peptides appeared to generally associate with more advanced cancer cases. We concluded that certain components of serum peptide signatures are common for different cancer signatures and putatively reflect general response of organism to the disease, yet other components of such signatures are more specific and most likely correspond to clinical stage of the malignancy.

Signal Partitioning Algorithm for Highly Efficient Gaussian Mixture Modeling in Mass Spectrometry

Mixture - modeling of mass spectra is an approach with many potential applications including peak detection and quantification, smoothing, de-noising, feature extraction and spectral signal compression. However, existing algorithms do not allow for automated analyses of whole spectra. Therefore, despite highlighting potential advantages of mixture modeling of mass spectra of peptide/protein mixtures and some preliminary results presented in several papers, the mixture modeling approach was so far not developed to the stage enabling systematic comparisons with existing software packages for proteomic mass spectra analyses. In this paper we present an efficient algorithm for Gaussian mixture modeling of proteomic mass spectra of different types (e.g., MALDI-ToF profiling, MALDI-IMS). The main idea is automated partitioning of protein mass spectral signal into fragments. The obtained fragments are separately decomposed into Gaussian mixture models. The parameters of the mixture models of fragments are then aggregated to form the mixture model of the whole spectrum. We compare the elaborated algorithm to existing algorithms for peak detection and we demonstrate improvements of peak detection efficiency obtained by using Gaussian mixture modeling. We also show applications of the elaborated algorithm to real proteomic datasets of low and high resolution.

Sources of High Variance between Probe Signals in Affymetrix Short Oligonucleotide Microarrays

High density oligonucleotide microarrays present a big challenge for statistical data processing methods which aim to separate changes induced by experimental factors from those caused by artifacts and measurement inaccuracies. Despite huge advances in the field of microarray probe design methods, the signal variation between probes that target a single transcript is substantially larger than their between-replicate array variability, suggesting a large influence of various probe-specific effects that introduce bias to the data. In this work we present the influence of probe-related design variations on the expression intensities of individual probes, focusing on five potential sources of high probe signal variance: the GC composition of the probe, the distance between individual probe target sites, G-quadruplex formation in the probe sequence, the occurrence of sequence motifs complementary to the oligo(dT) primer, and the specificity of unrecognized alternative splicing probeset assignment. By focusing on two high quality microarray datasets based on two distinct array designs we show the extent of variance between probes that target a specific transcript providing guidelines for the future design of microarrays and data processing methods.

Association between plasma proteome profiles analysed by mass spectrometry, a lymphocyte-based DNA-break repair assay and radiotherapy-induced acute mucosal reaction in head and neck cancer patients.

Purpose: The plasma proteome was analysed as a potential source of markers of radiosensitivity in patients treated with definitive radiotherapy for head and neck cancer. Materials and methods: Acute mucosal reactions that developed during radiotherapy were assessed in 55 patients. Blood samples were collected from each patient before the treatment and also from 50 healthy donors. The low-molecular-weight fraction of the plasma proteome (2,000–10,000 Da range) was analysed by the Matrix-Assisted Laser Desorption Ionisation mass spectrometry. The capacity for DNA break repair was assessed by the comet assay using lymphocytes irradiated in vitro. Results: Spectral components registered in plasma samples were used to build classifiers that discriminated patients from healthy individuals with about 90% specificity and sensitivity (components of 4469, 6929 and 8937 Da were the most essential for cancer classification). Four spectral components were identified (2219, 2454, 3431 and 5308 Da) whose abundances correlated with a maximal intensity of the acute reaction. Several spectral components whose abundances correlated with the rate of DNA repair in irradiated lymphocytes were also detected. Additionally, a more rapid escalation of an acute reaction was correlated with a higher level of unrepaired damage assessed by the comet assay. Conclusions: The plasma proteome could be considered as a potential source of predictive markers of acute reaction in patients with head and neck cancer treated with radiotherapy.

Radiation-induced changes in serum lipidome of head and neck cancer patients

Cancer radiotherapy (RT) induces response of the whole patient’s body that could be detected at the blood level. We aimed to identify changes induced in serum lipidome during RT and characterize their association with doses and volumes of irradiated tissue. Sixty-six patients treated with conformal RT because of head and neck cancer were enrolled in the study. Blood samples were collected before, during and about one month after the end of RT. Lipid extracts were analyzed using MALDI-oa-ToF mass spectrometry in positive ionization mode. The major changes were observed when pre-treatment and within-treatment samples were compared. Levels of several identified phosphatidylcholines, including (PC34), (PC36) and (PC38) variants, and lysophosphatidylcholines, including (LPC16) and (LPC18) variants, were first significantly decreased and then increased in post-treatment samples. Intensities of changes were correlated with doses of radiation received by patients. Of note, such correlations were more frequent when low-to-medium doses of radiation delivered during conformal RT to large volumes of normal tissues were analyzed. Additionally, some radiation-induced changes in serum lipidome were associated with toxicity of the treatment. Obtained results indicated the involvement of choline-related signaling and potential biological importance of exposure to clinically low/medium doses of radiation in patient’s body response to radiation.

Serum mass profile signature as a biomarker of early lung cancer

OBJECTIVES Circulating molecular biomarkers of lung cancer may allow the pre-selection of candidates for computed tomography screening or increase its efficacy. We aimed to identify features of serum mass profile distinguishing individuals with early lung cancer from healthy participants of the lung cancer screening program. METHODS Blood samples were collected during a low-dose computed tomography (LD-CT) screening program performed by one institution (Medical University of Gdansk, Poland). MALDI-ToF mass spectrometry was used to characterize the low-molecular-weight (1000-14,000Da) serum fraction. The analysis comprised 95 patients with early stage lung cancer (including 30 screen-detected cases) and a matched group of 285 healthy controls. The cases were split into two independent cohorts (discovery and validation), analyzed separately 6 months apart. RESULTS Several molecular components of serum (putatively components of endogenous peptidome) discriminating patients with early lung cancer from controls were identified in a discovery cohort. This allowed building an effective cancer classifier as a model tuned to maximize negative predictive value, with an area under the curve (AUC) of 0.88, a negative predictive value of 100%, and a positive predictive value of 48%. However, the classifier performed worse in a validation cohort including independent sample sets (AUC 0.73, NPV 88% and PPV 30%). CONCLUSIONS We developed a serum mass profile-based signature identifying patients with early lung cancer. Although this marker has insufficient value as a stand-alone preselecting tool for LD-CT screening, its potential clinical usefulness in evaluation of indeterminate pulmonary nodules deserves further investigation.

Influence of genetic background and oxidative stress response on risk of mandibular osteoradionecrosis after radiotherapy of head and neck cancer.

Osteoradionecrosis (ORN) of the mandible is a severe complication of head and neck radiotherapy (RT) treatment, where the impact of individual radiosensitivity has been a suggested explanation.

Serum lipid profile discriminates patients with early lung cancer from healthy controls

OBJECTIVES The role of a low-dose computed tomography lung cancer screening remains a matter of controversy due to its low specificity and high costs. Screening complementation with blood-based biomarkers may allow a more efficient pre-selection of candidates for imaging tests or discrimination between benign and malignant chest abnormalities detected by low-dose computed tomography (LD-CT). We searched for a molecular signature based on a serum lipid profile distinguishing individuals with early lung cancer from healthy participants of the lung cancer screening program. MATERIALS AND METHODS Blood samples were collected from 100 patients with early stage lung cancer (including 31 screen-detected cases) and from a matched group of 300 healthy participants of the lung cancer screening program. MALDI-ToF mass spectrometry was used to analyze the molecular profile of lipid-containing organic extract of serum samples in the 320-1000Da range. RESULTS Several components of the serum lipidome were detected, with abundances discriminating patients with early lung cancer from high-risk smokers. An effective cancer classifier was built with an area under the curve of 0.88. Corresponding negative predictive value was 98% and a positive predictive value was 42% when the classifier was tuned for maximum negative predictive value. Furthermore, the downregulation of a few lysophosphatidylcholines (LPC18:2, LPC18:1 and LPC18:0) in samples from cancer patients was confirmed using a complementary LC-MS approach (a reasonable cancer discrimination was possible based on LPC18:2 alone with 25% total weighted error of classification). CONCLUSIONS Lipid-based serum signature showed potential usefulness in discriminating early lung cancer patients from healthy individuals.

Stage I non-small-cell lung cancer: long-term results of lobectomy versus sublobar resection from the Polish National Lung Cancer Registry†

OBJECTIVES Anatomical lobar resection and mediastinal lymphadenectomy remain the standard for the treatment of early stage non-small-cell lung cancer (NSCLC) and are preferred over procedures such as segmentectomy or wedge resection. However, there is an ongoing debate concerning the influence of the extent of the resection on overall survival. The aim of this article was to assess the overall survival for different types of resection for Stage I NSCLC. METHODS We performed a retrospective analysis of the results of the surgical treatment of Stage I NSCLC. Between 1 January 2007 and 31 December 2013, the data from 6905 patients who underwent Stage I NSCLC operations were collected in the Polish National Lung Cancer Registry (PNLCR) and overall survival was assessed. A propensity score-matched analysis was used to compare 3 groups of patients, each consisting of 231 patients who underwent lobectomy, segmentectomy, or wedge resection. RESULTS In the unmatched and matched patient groups, lobectomy and segmentectomy were associated with a significant benefit compared to wedge resection regarding overall survival (log-rank P  < 0.001 and P  = 0.001). The Cox proportional hazard ratio comparing segmentectomy and lobectomy to wedge resection was 0.54 [95% confidence interval (CI): 0.37-0.77) and 0.44 (95% CI: 0.38-0.50), respectively, indicating a significant improvement in survival. There was no difference in the 5-year survival of patients after lobectomy (79.1%; 95% CI: 77.7-80.4%) or segmentectomy (78.3%; 95% CI: 70.6-86.0%). The 30-day mortality rate was 1.6, 2.6 and 1.4% for lobectomy, segmentectomy and wedge resection, respectively. Wedge resection was associated with a significantly lower 5-year survival rate (58.1%; 95% CI: 53.6-62.5%) compared to segmentectomy (78.3%; 95% CI: 70.6-86.0%) and lobectomy (79.1%; 95% CI: 77.7-80.5%). The propensity score matched analysis confirmed most of the results of the comparisons of unmatched study groups. CONCLUSIONS Wedge resection was associated with significantly lower 3-year and 5-year survival rates compared to the other methods of resection. There was no significant difference in 3-year or 5-year survival rates between lobectomy and segmentectomy. Segmentectomy, but not wedge resection, could be considered an alternative to lobectomy in the treatment of patients with Stage I NSCLC.