Proteomic strategies and challenges in tumor metastasis research. Clin Exp Metastasis. 2010;27:441-451. [PMID: 20607365 DOI: 10.1007/s10585-010-9339-7]

DEEPSMP: A deep learning model for predicting the ectodomain shedding events of membrane proteins

Membrane proteins play essential roles in modern medicine. In recent studies, some membrane proteins involved in ectodomain shedding events have been reported as the potential drug targets and biomarkers of some serious diseases. However, there are few effective tools for identifying the shedding event of membrane proteins. So, it is necessary to design an effective tool for predicting shedding event of membrane proteins. In this study, we design an end-to-end prediction model using deep neural networks with long short-term memory (LSTM) units and attention mechanism, to predict the ectodomain shedding events of membrane proteins only by sequence information. Firstly, the evolutional profiles are encoded from original sequences of these proteins by Position-Specific Iterated BLAST (PSI-BLAST) on Uniref50 database. Then, the LSTM units which contain memory cells are used to hold information from past inputs to the network and the attention mechanism is applied to detect sorting signals in proteins regardless of their position in the sequence. Finally, a fully connected dense layer and a softmax layer are used to obtain the final prediction results. Additionally, we also try to reduce overfitting of the model by using dropout, L2 regularization, and bagging ensemble learning in the model training process. In order to ensure the fairness of performance comparison, firstly we use cross validation process on training dataset obtained from an existing paper. The average accuracy and area under a receiver operating characteristic curve (AUC) of five-fold cross-validation are 81.19% and 0.835 using our proposed model, compared to 75% and 0.78 by a previously published tool, respectively. To better validate the performance of the proposed model, we also evaluate the performance of the proposed model on independent test dataset. The accuracy, sensitivity, and specificity are 83.14%, 84.08%, and 81.63% using our proposed model, compared to 70.20%, 71.97%, and 67.35% by the existing model. The experimental results validate that the proposed model can be regarded as a general tool for predicting ectodomain shedding events of membrane proteins. The pipeline of the model and prediction results can be accessed at the following URL: http://www.csbg-jlu.info/DeepSMP/.

The application of proteomics in different aspects of hepatocellular carcinoma research

Hepatocellular carcinoma (HCC) is one of the most common malignant tumors, which is causing the second leading cancer-related death worldwide. With the significant advances of high-throughput protein analysis techniques, the proteomics offered an extremely useful and versatile analytical platform for biomedical researches. In recent years, different proteomic strategies have been widely applied in the various aspects of HCC studies, ranging from screening the early diagnostic and prognostic biomarkers to in-depth investigating the underlying molecular mechanisms. In this review, we would like to systematically summarize the current applications of proteomics in hepatocellular carcinoma study, and discuss the challenges of applying proteomics in study clinical samples, as well as discuss the possible application of proteomics in precision medicine.

BIOLOGICAL SIGNIFICANCE

In this review, we have systematically summarized the current applications of proteomics in hepatocellular carcinoma study, ranging from screening biomarkers to in-depth investigating the underlying molecular mechanisms. In addition, we have discussed the challenges of applying proteomics in study clinical samples, as well as the possible applications of proteomics in precision medicine. We believe that this review would help readers to be better familiar with the recent progresses of clinical proteomics, especially in the field of hepatocellular carcinoma research.

Challenges of deciphering gastric cancer heterogeneity

Gastric cancer is in decline in most developed countries; however, it still accounts for a notable fraction of global mortality and morbidity related to cancer. High-throughput methods are rapidly changing our view and understanding of the molecular basis of gastric carcinogenesis. Today, it is widely accepted that the molecular complexity and heterogeneity, both inter- and intra-tumour, of gastric adenocarcinomas present significant obstacles in elucidating specific biomarkers for early detection of the disease. Although genome-wide sequencing and gene expression studies have revealed the intricate nature of the molecular changes that occur in tumour landscapes, the collected data and results are complex and sometimes contradictory. Several aberrant molecules have already been tested in clinical trials, although their diagnostic and prognostic utilities have not been confirmed thus far. The gold standard for the detection of sporadic gastric cancer is still the gastric endoscopy, which is considered invasive. In addition, genome-wide association studies have confirmed that genetic variations are important contributors to increased cancer risk and could participate in the initiation of malignant transformation. This hypothesis could in part explain the late onset of sporadic gastric cancers. The elaborate interplay of polymorphic low penetrance genes and lifestyle and environmental risk factors requires additional research to decipher their relative impacts on tumorigenesis. The purpose of this article is to present details of the molecular heterogeneity of sporadic gastric cancers at the DNA, RNA, and proteome levels and to discuss issues relevant to the translation of basic research data to clinically valuable tools. The focus of this work is the identification of relevant molecular changes that could be detected non-invasively.

Comparative serum proteomic analysis of serum diagnosis proteins of colorectal cancer based on magnetic bead separation and maldi-tof mass spectrometry

BACKGROUND

At present, the diagnosis of colorectal cancer (CRC) requires a colorectal biopsy which is an invasive procedure. We undertook this pilot study to develop an alternative method and potential new biomarkers for diagnosis, and validated a set of well-integrated tools called ClinProt to investigate the serum peptidome in CRC patients.

METHODS

Fasting blood samples from 67 patients diagnosed with CRC by histological diagnosis, 55 patients diagnosed with colorectal adenoma by biopsy, and 65 healthy volunteers were collected. Division was into a model construction group and an external validation group randomly. The present work focused on serum proteomic analysis of model construction group by ClinProt Kit combined with mass spectrometry. This approach allowed construction of a peptide pattern able to differentiate the studied populations. An external validation group was used to verify the diagnostic capability of the peptidome pattern blindly. An immunoassay method was used to determine serum CEA of CRC and controls.

RESULTS

The results showed 59 differential peptide peaks in CRC, colorectal adenoma and health volunteers. A genetic algorithm was used to set up the classification models. Four of the identified peaks at m/z 797, 810, 4078 and 5343 were used to construct peptidome patterns, achieving an accuracy of 100% (> CEA, P < 0. 05). Furthermore, the peptidome patterns could differentiate the validation group with high accuracy close to 100%.

CONCLUSIONS

Our results showed that proteomic analysis of serum with MALDI-TOF MS is a fast and reproducible approach, which may provide a novel approach to screening for CRC.

Fibrinogen alpha chain precursor and apolipoprotein A-I in urine as biomarkers for noninvasive diagnosis of calcium oxalate nephrolithiasis: a proteomics study

Calcium oxalate nephrolithiasis is the most common urological disease, but noninvasive and convenient methods of diagnosis are rarely available. Objective. The present study aimed to identify potential urine biomarkers for noninvasive diagnosis of CaOx nephrolithiasis. Methodology. Urine samples from 72 patients with CaOx nephrolithiasis and 30 healthy controls were collected and proteomics analysis was performed using matrix-assisted laser desorption/ionization-time of flight-mass spectrometer (MALDI-TOF-MS). Results. Thirteen proteins/peptides displayed statistically significant differences. The peptides of m/z 1207.23 and 2773.86 were selected by the genetic algorithm (GA) to build a possible diagnostic model. The area under the curve of m/z 1207.23 and 2773.86 was 0.936 and 0.987, respectively. The diagnostic model in distinguishing patients and healthy subjects showed 100% sensitivity and specificity. The peak at m/z 2773.86 was identified as fibrinogen alpha chain (FGA) with the sequence G.EGDFLAEGGGVR.G, and the peak at m/z 2773.86 was identified as apolipoprotein A-I (apoA-I) with the sequence L.PVLESFKVSFLSALEEYTKKLNTQ. Conclusion. The study results strongly suggested that urinary FGA and apoA-I are highly sensitive and specific biomarkers for noninvasive diagnosis of CaOx nephrolithiasis.

Quantitative proteomics analysis of early recurrence/metastasis of huge hepatocellular carcinoma following radical resection

BACKGROUND

Hepatic resection is the preferred treatment for huge hepatocellular carcinoma (>10 cm in diameter; H-HCC). However, the patients with H-HCC suffer from poor prognosis due to the early recurrence/metastasis. The underlying mechanism of H-HCC's early recurrence/metastasis is currently not well understood.

RESULTS

Here, we describe an Isobaric Tags for relative and absolute quantification (iTRAQ)-based quantitative proteomics approach to analyze the early recurrence/metastasis related proteins of H-HCC after radical resection through multidimensional chromatography coupled with tandem mass spectrometry (2DLC-MS/MS). The different protein expression profiles between the early recurrence/metastasis within 6 months(R/M≤6months) and late recurrence/metastasis within 6-12 months after surgery (R/M6-12months) were confirmed and might reveal different underlying molecular mechanisms. We identified 44 and 49 significantly differentially expressed proteins in the R/M≤6months group and the R/M6-12months group compared to the group who had no recurrence within 2 years post surgery (the NR/M group), respectively. Moreover, among those proteins, S100A12 and AMACR were down regulated in the R/M≤6months group but up-regulated in the R/M6-12months group; and this regulation was further confirmed in mRNA and protein level by Q-PCR, Western-Blot and Immunohistochemistry (IHC).

CONCLUSIONS

This current study presents the first proteomic profile of the early recurrence/metastasis of H-HCC. The results suggest that S100A12 and AMACR might be potential prognostic markers for predicting the early recurrence/metastasis of H-HCC after hepatectomy.

Identification alpha-2-HS-glycoprotein precursor and tubulin beta chain as serology diagnosis biomarker of colorectal cancer

BACKGROUND

Colorectal cancer (CRC) remains a major worldwide cause of cancer-related morbidity and mortality largely due to the insidious onset of the disease. The current clinical procedures utilized for disease diagnosis are invasive, unpleasant, and inconvenient. Hence, the need for simple blood tests that could be used for the early detection is crucial for its ultimate control and prevention.

METHODS

The present work is a case-control study focused on proteomic analysis of serum of healthy volunteers and CRC patients by the ClinProt profiling technology based on mass spectrometry. This approach allowed to identifying a pattern of proteins/peptides able to differentiate the studied populations. Moreover, some of peptides differentially expressed in the serum of patients as compared to healthy volunteers were identified by LTQ Orbitrap XL.

RESULTS

A Quick Classifier Algorithm was used to construct the peptidome patterns (m/z 1208, 1467, 1505, 1618, 1656 and 4215) for the identification of CRC from healthy volunteers with accuracy close to 100% (>CEA, P < 0.05). Peaks at m/z 1505 and 1618 were identified as alpha-2-HS-glycoprotein precursor and tubulin beta chain, respectively.

CONCLUSIONS

Alpha-2-HS-glycoprotein precursor and tubulin beta chain could be involved in the pathogenesis of CRC and perform as potential serology diagnosis biomarker.

VIRTUAL SLIDES

The virtual slide(s) for this article can be found here: http://www.diagnosticpathology.diagnomx.eu/vs/4796578761089186.

Tumor-induced perturbations of cytokines and immune cell networks

Until recently, the intrinsically high level of cross-talk between immune cells, the complexity of immune cell development, and the pleiotropic nature of cytokine signaling have hampered progress in understanding the mechanisms of immunosuppression by which tumor cells circumvent native and adaptive immune responses. One technology that has helped to shed light on this complex signaling network is the cytokine antibody array, which facilitates simultaneous screening of dozens to hundreds of secreted signal proteins in complex biological samples. The combined applications of traditional methods of molecular and cell biology with the high-content, high-throughput screening capabilities of cytokine antibody arrays and other multiplexed immunoassays have revealed a complex mechanism that involves multiple cytokine signals contributed not just by tumor cells but by stromal cells and a wide spectrum of immune cell types. This review will summarize the interactions among cancerous and immune cell types, as well as the key cytokine signals that are required for tumors to survive immunoediting in a dormant state or to grow and spread by escaping it. Additionally, it will present examples of how probing secreted cell-cell signal networks in the tumor microenvironment (TME) with cytokine screens have contributed to our current understanding of these processes and discuss the implications of this understanding to antitumor therapies.

Proteomic approaches in biomarker discovery: new perspectives in cancer diagnostics

Despite remarkable progress in proteomic methods, including improved detection limits and sensitivity, these methods have not yet been established in routine clinical practice. The main limitations, which prevent their integration into clinics, are high cost of equipment, the need for highly trained personnel, and last, but not least, the establishment of reliable and accurate protein biomarkers or panels of protein biomarkers for detection of neoplasms. Furthermore, the complexity and heterogeneity of most solid tumours present obstacles in the discovery of specific protein signatures, which could be used for early detection of cancers, for prediction of disease outcome, and for determining the response to specific therapies. However, cancer proteome, as the end-point of pathological processes that underlie cancer development and progression, could represent an important source for the discovery of new biomarkers and molecular targets for tailored therapies.

Tubulin beta chain, filamin A alpha isoform 1, and cytochrome b-c1 complex subunit 1 as serological diagnostic biomarkers of esophageal squamous cell carcinoma: a proteomics study

Despite the major advances in diagnosis and treatment, esophageal squamous cell carcinoma (ESCC) remains a major life-threatening disease. Early diagnosis is critical for guiding the therapeutic management of ESCC. This case-control study focused on the proteomic analysis of serum of healthy volunteers and ESCC patients using the ClinProt profiling technology based on mass spectrometry. A total of 80 healthy volunteers and 119 ESCC patients were enrolled. We identified a pattern of proteins/peptides (including m/z 1867, 2700, and 2094) and differentiated ESCC patients from healthy volunteers with sensitivity and specificity close to 100%. Using mass spectrometry (LTQ orbitrap XL), tubulin beta chain, filamin A alpha isoform 1, and cytochrome b-c1 complex subunit 1 were identified as the three differentially expressed proteins/peptides in the patient serum. These three dysregulated proteins/peptides could be involved in the pathogenesis of ESCC and may serve as putative serological diagnostic biomarkers of ESCC. We suggest that further proteomics and multi-omics research are warranted to identify novel post-genomics diagnostics that can in the future pave the way for personalized medicine for patients with ESCC, a cancer for which we currently lack an integrated battery of diagnostics in the field of oncology.

Towards a "Lyon molecular signature" to individualize the treatment of rectal cancer. Prognostic analysis of a prospective cohort of 94 rectal cancers T1-2-3 Nx MO to be the basis of a molecular signature

PURPOSE

In 1998 a translational research was initiated in Lyon aiming at identifying a prognostic "biomolecular signature" in rectal cancer. This paper presents the clinical outcome of the patients included in this study.

PATIENTS AND METHODS

A total of 94 patients were included between 1998 and 2001. A staging with rectoscopy and biopsies was performed before treatment. In case of surgery, the operative specimen was analysed to evaluate the pathological response. There were two types of treatment: neoadjuvant radiotherapy (with or without concurrent chemotherapy) followed by surgery (76 cases) and radiotherapy alone with 'contactherapy' often associated with external beam radiotherapy (18 patients).

RESULTS

The patients had a mean age of 63years. Stage was T1: 4, T2: 24, T3: 65 and T4: 1. The overall survival of the 94 patients was 62% at 8years with a rate of distant metastases of 29%. Rate of local recurrence at 8years was 6% in the neoadjuvant group and 16% in the radiotherapy group with an overall 8years survival in both groups respectively: 64% and 53%. There was a trend towards more metastases in cT3, tumour diameter above 4cm, circumferential extension. There was a significant increase in the risk of metastases for ypT3, ypN1-2 and Dworak score 1-2-3. In multivariate analysis ypT3 was significantly associated with a high rate of metastases (55%; P=0.0003).

CONCLUSION

The rate of distant metastases is a major prognostic factor. These clinical results will serve as the base line to identify a "biomolecular signature" which could complement the TN(M) classification.

Comparative serum proteomic analysis involving liver organ-specific metastasis-associated proteins of nasopharyngeal carcinoma

Metastasis is the main cause of cancer-related mortality; patients with liver metastases (LM) have the worst prognosis among patients with nasopharyngeal carcinoma (NPC). However, at present, few biomarkers for detecting organ-specific metastasis have been identified. Proteomics, an ultra-sensitive analytical technique, can detect molecular changes before organ-specific metastasis occurs. Analysis with matrix-assisted, laser desorption-ionization time-of-flight mass spectrometry (MALDI-TOF-MS), combined with magnetic chemical affinity beads is a new technique for evaluating protein separation. We sought to identify potential liver-specific, metastasis-associated proteomic printing in patients with NPC. We examined 64 serum samples from 50 patients who had pathologically confirmed NPC and 14 who had pathologically confirmed non-NPC with LM using MALDI-TOF-MS with weak cation bead protein chips. During follow-up of at least 37 months (maximum, 176 months) following radiotherapy, we confirmed 16 cases of LM (LM NPC), 16 cases without LM (non-LM NPC) and 18 cases without metastasis (non-M NPC). Using comparison analysis, 4 protein mass peaks, 4155.34, 4194.87, 4210.78 and 4249.56 m/z were identified as liver-specific, metastasis-associated protein peaks in NPC and two of them (4155 and 4249 m/z) met two different statistical criteria in both ClinProt software analyses and discriminant analyses. Models based on the 4 potential serum markers of NPC discriminated between LM NPC, non-LM NPC, non-M NPC and non-NPC LM analyzed with sieved markers. The recognition capability and cross-validation of these models for differentiating the above 4 groups are all approximately 80%. MALDI-TOF-MS combined with tree analysis models may provide a clinical diagnostic platform for detecting potential liver-specific, metastasis-associated proteomic printing in NPC. However, markedly differential proteins still need to be identified.

Maternal serum proteome changes between the first and third trimester of pregnancy in rural southern Nepal

Characterization of normal changes in the serum proteome during pregnancy may enhance understanding of maternal physiology and lead to the development of new gestational biomarkers. In 23 Nepalese pregnant women who delivered at term, two-dimensional difference in-gel electrophoresis (DIGE) was used to assess changes in relative protein abundance between paired serum samples collected in the first and third trimesters. One-hundred and forty-five of over 700 protein spots in DIGE gels (pI 4.2-6.8) exhibited nominally significant (p < 0.05) differences in abundance across trimesters. Additional filtering using a Bonferroni correction reduced the number of significant (p < 0.00019) spots to 61. Mass spectrometric analysis detected 38 proteins associated with gestational age, cytoskeletal remodeling, blood pressure regulation, lipid and nutrient transport, and inflammation. One new protein, pregnancy-specific β-glycoprotein 4 was detected. A follow-up isotope tagging for relative and absolute quantitation (iTRAQ) experiment of six mothers from the DIGE study revealed 111 proteins, of which 11 exhibited significant (p < 0.05) differences between trimesters. Four of these proteins: gelsolin, complement C1r subcomponent, α-1-acid glycoprotein, and α-1B-glycoprotein also changed in the DIGE analysis. Although not previously associated with normal pregnancy, gelsolin decreased in abundance by the third trimester (p < 0.01) in DIGE, iTRAQ and Western analyses. Changes in abundance of proteins in serum that are associated with syncytiotrophoblasts (gelsolin, pregnancy-specific β-1 glycoprotein 1 and β-2-glycoprotein I) probably reflect dynamics of a placental proteome shed into maternal circulation during pregnancy. Measurement of changes in the maternal serum proteome, when linked with birth outcomes, may yield biomarkers for tracking reproductive health in resource poor settings in future studies.

A large, consistent plasma proteomics data set from prospectively collected breast cancer patient and healthy volunteer samples

Background

Variability of plasma sample collection and of proteomics technology platforms has been detrimental to generation of large proteomic profile datasets from human biospecimens.

Methods

We carried out a clinical trial-like protocol to standardize collection of plasma from 204 healthy and 216 breast cancer patient volunteers. The breast cancer patients provided follow up samples at 3 month intervals. We generated proteomics profiles from these samples with a stable and reproducible platform for differential proteomics that employs a highly consistent nanofabricated ChipCube™ chromatography system for peptide detection and quantification with fast, single dimension mass spectrometry (LC-MS). Protein identification is achieved with subsequent LC-MS/MS analysis employing the same ChipCube™ chromatography system.

Results

With this consistent platform, over 800 LC-MS plasma proteomic profiles from prospectively collected samples of 420 individuals were obtained. Using a web-based data analysis pipeline for LC-MS profiling data, analyses of all peptide peaks from these plasma LC-MS profiles reveals an average coefficient of variability of less than 15%. Protein identification of peptide peaks of interest has been achieved with subsequent LC-MS/MS analyses and by referring to a spectral library created from about 150 discrete LC-MS/MS runs. Verification of peptide quantity and identity is demonstrated with several Multiple Reaction Monitoring analyses. These plasma proteomic profiles are publicly available through ProteomeCommons.

Conclusion

From a large prospective cohort of healthy and breast cancer patient volunteers and using a nano-fabricated chromatography system, a consistent LC-MS proteomics dataset has been generated that includes more than 800 discrete human plasma profiles. This large proteomics dataset provides an important resource in support of breast cancer biomarker discovery and validation efforts.