Support vector machines coupled with proteomics approaches for detecting biomarkers predicting chemotherapy resistance in small cell lung cancer

Study on risk factor analysis and model prediction of hyperuricemia in different populations

Objectives

The purpose of the present study was to explore the influencing factors of hyperuricemia (HUA) in different populations in Shandong Province based on clinical biochemical indicators. A prediction model for HUA was constructed to aid in the early prevention and screening of HUA.

Methods

In total, 705 cases were collected from five hospitals, and the risk factors were analyzed by Pearson correlation analysis, binary logistic regression, and receiver operating characteristic (ROC) curve in the gender and age groups. All data were divided into a training set and test set (7:3). The training set included age, gender, total protein (TP), low-density lipoprotein cholesterol (LDL-C), and 15 other indicators. The random forest (RF) and support vector machine (SVM) methods were used to build the HUA model, and model performances were evaluated through 10-fold cross-validation to select the optimal method. Finally, features were extracted, and the ROC curve of the test set was generated.

Results

TP, LDL-C, and glucose (GLU) were risk factors for HUA, and the area under the curve (AUC) value of the SVM validation set was 0.875.

Conclusion

The SVM model based on clinical biochemical indicators has good predictive ability for HUA, thus providing a reference for the diagnosis of HUA and the development of an HUA prediction model.

Correlation between small-cell lung cancer serum protein/peptides determined by matrix-assisted laser desorption/ionization time-of-flight mass spectrometry and chemotherapy efficacy

BACKGROUND

Currently, no effective measures are available to predict the curative efficacy of small-cell lung cancer (SCLC) chemotherapy. We expect to develop a method for effectively predicting the SCLC chemotherapy efficacy and prognosis in clinical practice in order to offer more pertinent therapeutic protocols for individual patients.

METHODS

We adopted matrix-assisted laser desorption/ionization time-of-flight mass spectrometry (MALDI-TOF-MS) and ClinPro Tools system to detect serum samples from 154 SCLC patients with different curative efficacy of standard chemotherapy and analyze the different peptides/proteins of SCLC patients to discover predictive tumor markers related to chemotherapy efficacy. Ten peptide/protein peaks were significantly different in the two groups.

RESULTS

A genetic algorithm model consisting of four peptides/proteins was developed from the training group to separate patients with different chemotherapy efficacies. Among them, three peptides/proteins (m/z 3323.35, 6649.03 and 6451.08) showed high expression in the disease progression group, whereas the peptide/protein at m/z 4283.18 was highly expressed in the disease response group. The classifier exhibited an accuracy of 91.4% (53/58) in the validation group. The survival analysis showed that the median progression-free survival (PFS) of 30 SCLC patients in disease response group was 9.0 months; in 28 cases in disease progression group, the median PFS was 3.0 months, a statistically significant difference (χ2 = 46.98, P < 0.001). The median overall survival (OS) of the two groups was 13.0 months and 7.0 months, a statistically significant difference (χ2 = 40.64, P < 0.001).

CONCLUSIONS

These peptides/proteins may be used as potential biological markers for prediction of the curative efficacy and prognosis for SCLC patients treated with standard regimen chemotherapy.

Proteomic Profiling of Small-Cell Lung Cancer: A Systematic Review

The accurate diagnosis of small-cell lung cancer (SCLC) is crucial, as treatment strategies differ from those of other lung cancers. This systematic review aims to identify proteins differentially expressed in SCLC compared to normal lung tissue, evaluating their potential utility in diagnosing and prognosing the disease. Additionally, the study identifies proteins differentially expressed between SCLC and large cell neuroendocrine carcinoma (LCNEC), aiming to discover biomarkers distinguishing between these two subtypes of neuroendocrine lung cancers. Following the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines, a comprehensive search was conducted across PubMed/MEDLINE, Scopus, Embase, and Web of Science databases. Studies reporting proteomics information and confirming SCLC and/or LCNEC through histopathological and/or cytopathological examination were included, while review articles, non-original articles, and studies based on animal samples or cell lines were excluded. The initial search yielded 1705 articles, and after deduplication and screening, 16 articles were deemed eligible. These studies revealed 117 unique proteins significantly differentially expressed in SCLC compared to normal lung tissue, along with 37 unique proteins differentially expressed between SCLC and LCNEC. In conclusion, this review highlights the potential of proteomics technology in identifying novel biomarkers for diagnosing SCLC, predicting its prognosis, and distinguishing it from LCNEC.

Surgical Methods and Social Factors Are Associated With Long-Term Survival in Follicular Thyroid Carcinoma: Construction and Validation of a Prognostic Model Based on Machine Learning Algorithms

Background

This study aimed to establish and verify an effective machine learning (ML) model to predict the prognosis of follicular thyroid cancer (FTC), and compare it with the eighth edition of the American Joint Committee on Cancer (AJCC) model.

Methods

Kaplan-Meier method and Cox regression model were used to analyze the risk factors of cancer-specific survival (CSS). Propensity-score matching (PSM) was used to adjust the confounding factors of different surgeries. Nine different ML algorithms,including eXtreme Gradient Boosting (XGBoost), Light Gradient Boosting Machine (LightGBM), Random Forests (RF), Logistic Regression (LR), Adaptive Boosting (AdaBoost), Gaussian Naive Bayes (GaussianNB), K-Nearest Neighbor (KNN), Support Vector Machine (SVM) and Multi-Layer Perceptron (MLP),were used to build prognostic models of FTC.10-fold cross-validation and SHapley Additive exPlanations were used to train and visualize the optimal ML model.The AJCC model was built by multivariate Cox regression and visualized through nomogram. The performance of the XGBoost model and AJCC model was mainly assessed using the area under the receiver operating characteristic (AUROC).

Results

Multivariate Cox regression showed that age, surgical methods, marital status, T classification, N classification and M classification were independent risk factors of CSS. Among different surgeries, the prognosis of one-sided thyroid lobectomy plus isthmectomy (LO plus IO) was the best, followed by total thyroidectomy (hazard ratios: One-sided thyroid LO plus IO, 0.086[95% confidence interval (CI),0.025-0.290], P&lt;0.001; total thyroidectomy (TT), 0.490[95%CI,0.295-0.814], P=0.006). PSM analysis proved that one-sided thyroid LO plus IO, TT, and partial thyroidectomy had no significant differences in long-term prognosis. Our study also revealed that married patients had better prognosis than single, widowed and separated patients (hazard ratios: single, 1.686[95%CI,1.146-2.479], P=0.008; widowed, 1.671[95%CI,1.163-2.402], P=0.006; separated, 4.306[95%CI,2.039-9.093], P&lt;0.001). Among different ML algorithms, the XGBoost model had the best performance, followed by Gaussian NB, RF, LR, MLP, LightGBM, AdaBoost, KNN and SVM. In predicting FTC prognosis, the predictive performance of the XGBoost model was relatively better than the AJCC model (AUROC: 0.886 vs. 0.814).

Conclusion

For high-risk groups, effective surgical methods and well marital status can improve the prognosis of FTC. Compared with the traditional AJCC model, the XGBoost model has relatively better prediction accuracy and clinical usage.

Big Data to Knowledge: Application of Machine Learning to Predictive Modeling of Therapeutic Response in Cancer

Background

In recent years, the availability of high throughput technologies, establishment of large molecular patient data repositories, and advancement in computing power and storage have allowed elucidation of complex mechanisms implicated in therapeutic response in cancer patients. The breadth and depth of such data, alongside experimental noise and missing values, requires a sophisticated human-machine interaction that would allow effective learning from complex data and accurate forecasting of future outcomes, ideally embedded in the core of machine learning design.

Objective

In this review, we will discuss machine learning techniques utilized for modeling of treatment response in cancer, including Random Forests, support vector machines, neural networks, and linear and logistic regression. We will overview their mathematical foundations and discuss their limitations and alternative approaches in light of their application to therapeutic response modeling in cancer.

Conclusion

We hypothesize that the increase in the number of patient profiles and potential temporal monitoring of patient data will define even more complex techniques, such as deep learning and causal analysis, as central players in therapeutic response modeling.

Development of Machine Learning Models to Validate a Medication Regimen Complexity Scoring Tool for Critically Ill Patients

INTRODUCTION

The Medication Regimen Complexity -Intensive Care Unit (MRC-ICU) is the first tool for measuring medication regimen complexity in critically ill patients. This study tested machine learning (ML) models to investigate the relationship between medication regimen complexity and patient outcomes.

METHODS

This study was a single-center, retrospective observational evaluation of 130 adults admitted to the medical ICU. The MRC-ICU score was utilized to improve the inpatient model's prediction accuracy. Three models were proposed: model I, demographic data without medication data; model II, demographic data and medication regimen complexity variables; and model III: demographic data and the MRC-ICU score. A total of 6 ML classifiers was developed: k-nearest neighbor (KNN), naïve Bayes (NB), random forest, support vector machine, neural network, and logistic classifier (LC). They were developed and tested using electronic health record data to predict inpatient mortality.

RESULTS

The results demonstrated that adding medication regimen complexity variables (model II) and the MRC-ICU score (model III) improved inpatient mortality prediction.. The LC outperformed the other classifiers (KNN and NB), with an overall accuracy of 83%, sensitivity (Se) of 87%, specificity of 67%, positive predictive value of 93%, and negative predictive value of 46%. The APACHE III score and the MRC-ICU score at the 24-hour interval were the 2 most important variables.

CONCLUSION AND RELEVANCE

Inclusion of the MRC-ICU score improved the prediction of patient outcomes on the previously established APACHE III score. This novel, proof-of-concept methodology shows promise for future application of the MRC-ICU scoring tool for patient outcome predictions.

Roles of S100 family members in drug resistance in tumors: Status and prospects

Chemotherapy and targeted therapy can significantly improve survival rates in cancer, but multiple drug resistance (MDR) limits the efficacy of these approaches. Understanding the molecular mechanisms underlying MDR is crucial for improving drug efficacy and clinical outcomes of patients with cancer. S100 proteins belong to a family of calcium-binding proteins and have various functions in tumor development. Increasing evidence demonstrates that the dysregulation of various S100 proteins contributes to the development of drug resistance in tumors, providing a basis for the development of predictive and prognostic biomarkers in cancer. Therefore, a combination of biological inhibitors or sensitizers of dysregulated S100 proteins could enhance therapeutic responses. In this review, we provide a detailed overview of the mechanisms by which S100 family members influence resistance of tumors to cancer treatment, with a focus on the development of effective strategies for overcoming MDR.

Proteomic biomarkers for lung cancer progression

Aim: Lung adenocarcinoma (LUAD) and lung squamous-cell carcinoma (LUSC) are two major subtypes of lung cancer and constitute about 70% of all the lung cancer cases. The patient's lifespan and living quality will be significantly improved if they are diagnosed at an early stage and adequately treated. Methods & results: This study comprehensively screened the proteomic dataset of both LUAD and LUSC, and proposed classification models for the progression stages of LUAD and LUSC with accuracies 86.51 and 89.47%, respectively. Discussion & conclusion: A comparative analysis was also carried out on related transcriptomic datasets, which indicates that the proposed biomarkers provide discerning power for accurate stage prediction, and will be improved when larger-scale proteomic quantitative technologies become available.

Mass Spectrometry Analyses of Multicellular Tumor Spheroids

Multicellular tumor spheroids (MCTS) are a powerful biological in vitro model, which closely mimics the 3D structure of primary avascularized tumors. Mass spectrometry (MS) has established itself as a powerful analytical tool, not only to better understand and describe the complex structure of MCTS, but also to monitor their response to cancer therapeutics. The first part of this review focuses on traditional mass spectrometry approaches with an emphasis on elucidating the molecular characteristics of these structures. Then the mass spectrometry imaging (MSI) approaches used to obtain spatially defined information from MCTS is described. Finally the analysis of primary spheroids, such as those present in ovarian cancer, and the great potential that mass spectrometry analysis of these structures has for improved understanding of cancer progression and for personalized in vitro therapeutic testing is discussed.

Applications of Support Vector Machine (SVM) Learning in Cancer Genomics

Machine learning with maximization (support) of separating margin (vector), called support vector machine (SVM) learning, is a powerful classification tool that has been used for cancer genomic classification or subtyping. Today, as advancements in high-throughput technologies lead to production of large amounts of genomic and epigenomic data, the classification feature of SVMs is expanding its use in cancer genomics, leading to the discovery of new biomarkers, new drug targets, and a better understanding of cancer driver genes. Herein we reviewed the recent progress of SVMs in cancer genomic studies. We intend to comprehend the strength of the SVM learning and its future perspective in cancer genomic applications.

Construction of a 26‑feature gene support vector machine classifier for smoking and non‑smoking lung adenocarcinoma sample classification

The present study aimed to identify the feature genes associated with smoking in lung adenocarcinoma (LAC) samples and explore the underlying mechanism. Three gene expression datasets of LAC samples were downloaded from the Gene Expression Omnibus database through pre-set criteria and the expression data were processed using meta-analysis. Differentially expressed genes (DEGs) between LAC samples of smokers and non-smokers were identified using limma package in R. The classification accuracy of selected DEGs were visualized using hierarchical clustering analysis in R language. A protein-protein interaction (PPI) network was constructed using gene interaction data from the Human Protein Reference Database for the DEGs. Betweenness centrality was calculated for each node in the network and genes with the greatest BC values were utilized for the construction of the support vector machine (SVM) classifier. The dataset GSE43458 was used as the training dataset for the construction and the other datasets (GSE12667 and GSE10072) were used as the validation datasets. The classification accuracy of the classifier was tested using sensitivity, specificity, positive predictive value, negative predictive value and area under curve parameters with the pROC package in R language. The feature genes in the SVM classifier were subjected to pathway enrichment analysis using Fisher's exact test. A total of 347 genes were identified to be differentially expressed between samples of smokers and non-smokers. The PPI network of DEGs were comprised of 202 nodes and 300 edges. An SVM classifier comprised of 26 feature genes was constructed to distinguish between different LAC samples, with prediction accuracies for the GSE43458, GSE12667 and GSE10072 datasets of 100, 100 and 94.83%, respectively. Furthermore, the 26 feature genes that were significantly enriched in 9 overrepresented biological pathways, including extracellular matrix-receptor interaction, proteoglycans in cancer, cell adhesion molecules, p53 signaling pathway, microRNAs in cancer and apoptosis, were identified to be smoking-related genes in LAC. In conclusion, an SVM classifier with a high prediction accuracy for smoking and non-smoking samples was obtained. The genes in the classifier may likely be the potential feature genes associated with the development of patients with LAC who smoke.

[Detection of Serum Peptides in Patients with Lung Squamous Cell Carcinoma by MALDI-TOF-MS and Analysis of Their Correlation with Chemotherapy Efficacy]

BACKGROUND

Treatment options for patients with squamous cell carcinoma of the lung (SCC) are limited in chemotherapy. However, not all patients could benefit form standard platinum regimen. Considering the dismal prognosis of patients with advanced SCC, a greater focus on selecting sensitive chemotherapy regimens remains of upmost importance to improve outcomes in this disease. In this study, we used matrix-assisted laser desorption/ionization time-of-flight mass spectrometry to detect pre-chemotherapy serum peptides in advanced lung squamous cell carcinoma patients accepting paclitaxel combined with platinum chemotherapy and to analyze the correlation between serum peptides and chemotherapy efficacy.

METHODS

Patients with advanced lung squamous cell carcinoma received paclitaxel combining with platinum chemotherapy and evaluated the efficacy every two cycles. Evaluation of complete response (CR) or partial response (PR) patients defined as sensitive group, progressive disease (PD) patients defined as resistant group. Serum samples were collected from patients with lung squamous cell carcinoma. Eighty-one patients were randomly divided into training group (sensitive group I and resistant group I) and validation group (sensitive group II and resistant group II) according to the ratio of 3:1. Serum samples were pretreated and Matrix-assisted laser desorption/ionization time-of-flight mass spectrometry (MALDI-TOF-MS) was used to detect serum peptide fingerprints. ClinProTools software was used to analyze the differences between the sensitive group I and the resistant group I. Three kinds of biological algorithms (SNN, GA, QC) built in CPT software were used to establish the curative effect prediction model respectively and the optimal algorithm was selected. The validation group was used for blind verification.

RESULTS

Thirty sensitive patients and 31 resistant patients were enrolled in the training group. Ten sensitive patients and 10 resistant patients were included in the validation group. The training group had 96 differentially expressed peptides in the sensitive and resistant patients, with 16 statistically significant peptides (P<0.001). The predictive model was established by 5 polypeptides (1,897.75 Da, 2,023.93 Da, 3,683.36 Da, 4,269.56 Da, 5,341.29 Da). The recognition rate of this model was 89.18% and the cross validation rate was 95.11%. The accuracy of the model was 85%, the sensitivity was 90.0% and the specificity was 80.0%. The median PFS in the sensitive group was better than patients in the resistant group (7.2 months 95%CI: 4.4-14.5 vs 1.8 months 95%CI: 0.7-3.5). The results showed that the differential peptides 4,232.04 Da and 4,269.56 Da were correlated with PFS in patients with lung squamous cell carcinoma (P<0.001).

CONCLUSIONS

MALDI-TOF-MS was used to detect the difference of serum peptides between sensitive and resistant groups. The preliminary curative effect prediction model was used to predict the efficacy of paclitaxel combined with platinum regimen. However, this model need further investigations to verify the accuracy and the sensitivity.

Label-Free LC-MS/MS Proteomic Analysis of Cerebrospinal Fluid Identifies Protein/Pathway Alterations and Candidate Biomarkers for Amyotrophic Lateral Sclerosis

Analysis of the cerebrospinal fluid (CSF) proteome has proven valuable to the study of neurodegenerative disorders. To identify new protein/pathway alterations and candidate biomarkers for amyotrophic lateral sclerosis (ALS), we performed comparative proteomic profiling of CSF from sporadic ALS (sALS), healthy control (HC), and other neurological disease (OND) subjects using label-free liquid chromatography-tandem mass spectrometry (LC-MS/MS). A total of 1712 CSF proteins were detected and relatively quantified by spectral counting. Levels of several proteins with diverse biological functions were significantly altered in sALS samples. Enrichment analysis was used to link these alterations to biological pathways, which were predominantly related to inflammation, neuronal activity, and extracellular matrix regulation. We then used our CSF proteomic profiles to create a support vector machines classifier capable of discriminating training set ALS from non-ALS (HC and OND) samples. Four classifier proteins, WD repeat-containing protein 63, amyloid-like protein 1, SPARC-like protein 1, and cell adhesion molecule 3, were identified by feature selection and externally validated. The resultant classifier distinguished ALS from non-ALS samples with 83% sensitivity and 100% specificity in an independent test set. Collectively, our results illustrate the utility of CSF proteomic profiling for identifying ALS protein/pathway alterations and candidate disease biomarkers.

Mass spectrometry-based serum and plasma peptidome profiling for prediction of treatment outcome in patients with solid malignancies

INTRODUCTION

Treatment selection tools are needed to enhance the efficacy of targeted treatment in patients with solid malignancies. Providing a readout of aberrant signaling pathways and proteolytic events, mass spectrometry-based (MS-based) peptidomics enables identification of predictive biomarkers, whereas the serum or plasma peptidome may provide easily accessible signatures associated with response to treatment. In this systematic review, we evaluate MS-based peptide profiling in blood for prompt clinical implementation.

METHODS

PubMed and Embase were searched for studies using a syntax based on the following hierarchy: (a) blood-based matrix-assisted or surface-enhanced laser desorption/ionization time-of-flight MS peptide profiling (b) in patients with solid malignancies (c) prior to initiation of any treatment modality, (d) with availability of outcome data.

RESULTS

Thirty-eight studies were eligible for review; the majority were performed in patients with non-small cell lung cancer (NSCLC). Median classification prediction accuracy was 80% (range: 66%-93%) in 11 models from 14 studies reporting an MS-based classification model. A pooled analysis of 9 NSCLC studies revealed clinically significant median progression-free survival in patients classified as "poor outcome" and "good outcome" of 2.0 ± 1.06 months and 4.6 ± 1.60 months, respectively; median overall survival was also clinically significant at 4.01 ± 1.60 months and 10.52 ± 3.49 months, respectively.

CONCLUSION

Pretreatment MS-based serum and plasma peptidomics have shown promising results for prediction of treatment outcome in patients with solid tumors. Limited sample sizes and absence of signature validation in many studies have prohibited clinical implementation thus far. Our pooled analysis and recent results from the PROSE study indicate that this profiling approach enables treatment selection, but additional prospective studies are warranted.

Systems biology of cisplatin resistance: past, present and future

The platinum derivative cis-diamminedichloroplatinum(II), best known as cisplatin, is currently employed for the clinical management of patients affected by testicular, ovarian, head and neck, colorectal, bladder and lung cancers. For a long time, the antineoplastic effects of cisplatin have been fully ascribed to its ability to generate unrepairable DNA lesions, hence inducing either a permanent proliferative arrest known as cellular senescence or the mitochondrial pathway of apoptosis. Accumulating evidence now suggests that the cytostatic and cytotoxic activity of cisplatin involves both a nuclear and a cytoplasmic component. Despite the unresolved issues regarding its mechanism of action, the administration of cisplatin is generally associated with high rates of clinical responses. However, in the vast majority of cases, malignant cells exposed to cisplatin activate a multipronged adaptive response that renders them less susceptible to the antiproliferative and cytotoxic effects of the drug, and eventually resume proliferation. Thus, a large fraction of cisplatin-treated patients is destined to experience therapeutic failure and tumor recurrence. Throughout the last four decades great efforts have been devoted to the characterization of the molecular mechanisms whereby neoplastic cells progressively lose their sensitivity to cisplatin. The advent of high-content and high-throughput screening technologies has accelerated the discovery of cell-intrinsic and cell-extrinsic pathways that may be targeted to prevent or reverse cisplatin resistance in cancer patients. Still, the multifactorial and redundant nature of this phenomenon poses a significant barrier against the identification of effective chemosensitization strategies. Here, we discuss recent systems biology studies aimed at deconvoluting the complex circuitries that underpin cisplatin resistance, and how their findings might drive the development of rational approaches to tackle this clinically relevant problem.

A quantitative system for discriminating induced pluripotent stem cells, embryonic stem cells and somatic cells

Induced pluripotent stem cells (iPSCs) derived from somatic cells (SCs) and embryonic stem cells (ESCs) provide promising resources for regenerative medicine and medical research, leading to a daily identification of new cell lines. However, an efficient system to discriminate the different types of cell lines is lacking. Here, we develop a quantitative system to discriminate the three cell types, iPSCs, ESCs, and SCs. The system consists of DNA-methylation biomarkers and mathematical models, including an artificial neural network and support vector machines. All biomarkers were unbiasedly selected by calculating an eigengene score derived from analysis of genome-wide DNA methylations. With 30 biomarkers, or even with as few as 3 top biomarkers, this system can discriminate SCs from pluripotent cells (PCs, including ESCs and iPSCs) with almost 100% accuracy. With approximately 100 biomarkers, the system can distinguish ESCs from iPSCs with an accuracy of 95%. This robust system performs precisely with raw data without normalization as well as with converted data in which the continuous methylation levels are accounted. Strikingly, this system can even accurately predict new samples generated from different microarray platforms and the next-generation sequencing. The subtypes of cells, such as female and male iPSCs and fetal and adult SCs, can also be discriminated with this method. Thus, this novel quantitative system works as an accurate framework for discriminating the three cell types, iPSCs, ESCs, and SCs. This strategy also supports the notion that DNA-methylation generally varies among the three cell types.