An individualized protein-based prognostic model to stratify pediatric patients with papillary thyroid carcinoma.

Pediatric papillary thyroid carcinomas (PPTCs) exhibit high inter-tumor heterogeneity and currently lack widely adopted recurrence risk stratification criteria. Hence, we propose a machine learning-based objective method to individually predict their recurrence risk. We retrospectively collect and evaluate the clinical factors and proteomes of 83 pediatric benign (PB), 85 pediatric malignant (PM) and 66 adult malignant (AM) nodules, and quantify 10,426 proteins by mass spectrometry. We find 243 and 121 significantly dysregulated proteins from PM vs. PB and PM vs. AM, respectively. Function and pathway analyses show the enhanced activation of the inflammatory and immune system in PM patients compared with the others. Nineteen proteins are selected to predict recurrence using a machine learning model with an accuracy of 88.24%. Our study generates a protein-based personalized prognostic prediction model that can stratify PPTC patients into high- or low-recurrence risk groups, providing a reference for clinical decision-making and individualized treatment.

Proteomic profiling of ovarian clear cell carcinomas identifies prognostic biomarkers for chemotherapy.

Clear cell ovarian carcinoma (CCOC) is a relatively rare subtype of ovarian cancer (OC) with high degree of resistance to standard chemotherapy. Little is known about the underlying molecular mechanisms, and it remains a challenge to predict its prognosis after chemotherapy. Here, we first analyzed the proteome of 35 formalin-fixed paraffin-embedded (FFPE) CCOC tissue specimens from a cohort of 32 patients with CCOC (H1 cohort) and characterized 8697 proteins using data-independent acquisition mass spectrometry (DIA-MS). We then performed proteomic analysis of 28 fresh frozen (FF) CCOC tissue specimens from an independent cohort of 24 patients with CCOC (H2 cohort), leading to the identification of 9409 proteins with DIA-MS. After bioinformatics analysis, we narrowed our focus to 15 proteins significantly correlated with the recurrence free survival (RFS) in both cohorts. These proteins are mainly involved in DNA damage response, extracellular matrix (ECM), and mitochondrial metabolism. Parallel reaction monitoring (PRM)-MS was adopted to validate the prognostic potential of the 15 proteins in the H1 cohort and an independent confirmation cohort (H3 cohort). Interferon-inducible transmembrane protein 1 (IFITM1) was observed as a robust prognostic marker for CCOC in both PRM data and immunohistochemistry (IHC) data. Taken together, this study presents a CCOC proteomic data resource and a single promising protein, IFITM1, which could potentially predict the recurrence and survival of CCOC.

Risk stratification of papillary thyroid cancers using multidimensional machine learning.

BACKGROUND: Papillary thyroid cancer (PTC) is one of the most common endocrine malignancies with different risk levels. However, preoperative risk assessment of PTC is still a challenge in the worldwide. Here, the authors first report a Preoperative Risk Assessment Classifier for PTC (PRAC-PTC) by multidimensional features including clinical indicators, immune indices, genetic feature, and proteomics. MATERIALS AND METHODS: The 558 patients collected from June 2013 to November 2020 were allocated to three groups: the discovery set [274 patients, 274 formalin-fixed paraffin-embedded (FFPE)], the retrospective test set (166 patients, 166 FFPE), and the prospective test set (118 patients, 118 fine-needle aspiration). Proteomic profiling was conducted by FFPE and fine-needle aspiration tissues from the patients. Preoperative clinical information and blood immunological indices were collected. The BRAFV600E mutation were detected by the amplification refractory mutation system. RESULTS: The authors developed a machine learning model of 17 variables based on the multidimensional features of 274 PTC patients from a retrospective cohort. The PRAC-PTC achieved areas under the curve (AUC) of 0.925 in the discovery set and was validated externally by blinded analyses in a retrospective cohort of 166 PTC patients (0.787 AUC) and a prospective cohort of 118 PTC patients (0.799 AUC) from two independent clinical centres. Meanwhile, the preoperative predictive risk effectiveness of clinicians was improved with the assistance of PRAC-PTC, and the accuracies reached at 84.4% (95% CI: 82.9-84.4) and 83.5% (95% CI: 82.2-84.2) in the retrospective and prospective test sets, respectively. CONCLUSION: This study demonstrated that the PRAC-PTC that integrating clinical data, gene mutation information, immune indices, high-throughput proteomics and machine learning technology in multicentre retrospective and prospective clinical cohorts can effectively stratify the preoperative risk of PTC and may decrease unnecessary surgery or overtreatment.

Identifying Patients With Rapid Progression From Hormone-Sensitive to Castration-Resistant Prostate Cancer: A Retrospective Study.

Prostate cancer (PCa) is the second most prevalent malignancy and the fifth cause of cancer-related deaths in men. A crucial challenge is identifying the population at risk of rapid progression from hormone-sensitive prostate cancer (HSPC) to lethal castration-resistant prostate cancer (CRPC). We collected 78 HSPC biopsies and measured their proteomes using pressure cycling technology and a pulsed data-independent acquisition pipeline. We quantified 7355 proteins using these HSPC biopsies. A total of 251 proteins showed differential expression between patients with a long- or short-term progression to CRPC. Using a random forest model, we identified seven proteins that significantly discriminated long- from short-term progression patients, which were used to classify PCa patients with an area under the curve of 0.873. Next, one clinical feature (Gleason sum) and two proteins (BGN and MAPK11) were found to be significantly associated with rapid disease progression. A nomogram model using these three features was generated for stratifying patients into groups with significant progression differences (p-value = 1.3×10-4). To conclude, we identified proteins associated with a fast progression to CRPC and an unfavorable prognosis. Based on these proteins, our machine learning and nomogram models stratified HSPC into high- and low-risk groups and predicted their prognoses. These models may aid clinicians in predicting the progression of patients, guiding individualized clinical management and decisions.

DPHL v.2: An updated and comprehensive DIA pan-human assay library for quantifying more than 14,000 proteins.

A comprehensive pan-human spectral library is critical for biomarker discovery using mass spectrometry (MS)-based proteomics. DPHL v.1, a previous pan-human library built from 1,096 data-dependent acquisition (DDA) MS data of 16 human tissue types, allows quantifying of 10,943 proteins. Here, we generated DPHL v.2 from 1,608 DDA-MS data. The data included 586 DDA-MS data acquired from 18 tissue types, while 1,022 files were derived from DPHL v.1. DPHL v.2 thus comprises data from 24 sample types, including several cancer types (lung, breast, kidney, and prostate cancer, among others). We generated four variants of DPHL v.2 to include semi-tryptic peptides and protein isoforms. DPHL v.2 was then applied to two colorectal cancer cohorts. The numbers of identified and significantly dysregulated proteins increased by at least 21.7% and 14.2%, respectively, compared with DPHL v.1. Our findings show that the increased human proteome coverage of DPHL v.2 provides larger pools of potential protein biomarkers.

EVNet: An Explainable Deep Network for Dimension Reduction.

Dimension reduction (DR) is commonly utilized to capture the intrinsic structure and transform high-dimensional data into low-dimensional space while retaining meaningful properties of the original data. It is used in various applications, such as image recognition, single-cell sequencing analysis, and biomarker discovery. However, contemporary parametric-free and parametric DR techniques suffer from several significant shortcomings, such as the inability to preserve global and local features and the pool generalization performance. On the other hand, regarding explainability, it is crucial to comprehend the embedding process, especially the contribution of each part to the embedding process, while understanding how each feature affects the embedding results that identify critical components and help diagnose the embedding process. To address these problems, we have developed a deep neural network method called EVNet, which provides not only excellent performance in structural maintainability but also explainability to the DR therein. EVNet starts with data augmentation and a manifold-based loss function to improve embedding performance. The explanation is based on saliency maps and aims to examine the trained EVNet parameters and contributions of components during the embedding process. The proposed techniques are integrated with a visual interface to help the user to adjust EVNet to achieve better DR performance and explainability. The interactive visual interface makes it easier to illustrate the data features, compare different DR techniques, and investigate DR. An in-depth experimental comparison shows that EVNet consistently outperforms the state-of-the-art methods in both performance measures and explainability.

Integrated proteogenomic characterization of medullary thyroid carcinoma.

Medullary thyroid carcinoma (MTC) is a rare neuroendocrine malignancy derived from parafollicular cells (C cells) of the thyroid. Here we presented a comprehensive multi-omics landscape of 102 MTCs through whole-exome sequencing, RNA sequencing, DNA methylation array, proteomic and phosphoproteomic profiling. Integrated analyses identified BRAF and NF1 as novel driver genes in addition to the well-characterized RET and RAS proto-oncogenes. Proteome-based stratification of MTCs revealed three molecularly heterogeneous subtypes named as: (1) Metabolic, (2) Basal and (3) Mesenchymal, which are distinct in genetic drivers, epigenetic modification profiles, clinicopathologic factors and clinical outcomes. Furthermore, we explored putative therapeutic targets of each proteomic subtype, and found that two tenascin family members TNC/TNXB might serve as potential prognostic biomarkers for MTC. Collectively, our study expands the knowledge of MTC biology and therapeutic vulnerabilities, which may serve as an important resource for future investigation on this malignancy.

Spatially resolved proteomics via tissue expansion.

Spatially resolved proteomics is an emerging approach for mapping proteome heterogeneity of biological samples, however, it remains technically challenging due to the complexity of the tissue microsampling techniques and mass spectrometry analysis of nanoscale specimen volumes. Here, we describe a spatially resolved proteomics method based on the combination of tissue expansion with mass spectrometry-based proteomics, which we call Expansion Proteomics (ProteomEx). ProteomEx enables quantitative profiling of the spatial variability of the proteome in mammalian tissues at ~160 µm lateral resolution, equivalent to the tissue volume of 0.61 nL, using manual microsampling without the need for custom or special equipment. We validated and demonstrated the utility of ProteomEx for streamlined large-scale proteomics profiling of biological tissues including brain, liver, and breast cancer. We further applied ProteomEx for identifying proteins associated with Alzheimer's disease in a mouse model by comparative proteomic analysis of brain subregions.

Comparative Evaluation of Proteome Discoverer and FragPipe for the TMT-Based Proteome Quantification.

Isobaric labeling-based proteomics is widely applied in deep proteome quantification. Among the platforms for isobaric labeled proteomic data analysis, the commercial software Proteome Discoverer (PD) is widely used, incorporating the search engine CHIMERYS, while FragPipe (FP) is relatively new, free for noncommercial purposes, and integrates the engine MSFragger. Here, we compared PD and FP over three public proteomic data sets labeled using 6plex, 10plex, and 16plex tandem mass tags. Our results showed the protein abundances generated by the two software are highly correlated. PD quantified more proteins (10.02%, 15.44%, 8.19%) than FP with comparable NA ratios (0.00% vs. 0.00%, 0.85% vs. 0.38%, and 11.74% vs. 10.52%) in the three data sets. Using the 16plex data set, PD and FP outputs showed high consistency in quantifying technical replicates, batch effects, and functional enrichment in differentially expressed proteins. However, FP saved 93.93%, 96.65%, and 96.41% of processing time compared to PD for analyzing the three data sets, respectively. In conclusion, while PD is a well-maintained commercial software integrating various additional functions and can quantify more proteins, FP is freely available and achieves similar output with a shorter computational time. Our results will guide users in choosing the most suitable quantification software for their needs.

Artificial intelligence defines protein-based classification of thyroid nodules.

Determination of malignancy in thyroid nodules remains a major diagnostic challenge. Here we report the feasibility and clinical utility of developing an AI-defined protein-based biomarker panel for diagnostic classification of thyroid nodules: based initially on formalin-fixed paraffin-embedded (FFPE), and further refined for fine-needle aspiration (FNA) tissue specimens of minute amounts which pose technical challenges for other methods. We first developed a neural network model of 19 protein biomarkers based on the proteomes of 1724 FFPE thyroid tissue samples from a retrospective cohort. This classifier achieved over 91% accuracy in the discovery set for classifying malignant thyroid nodules. The classifier was externally validated by blinded analyses in a retrospective cohort of 288 nodules (89% accuracy; FFPE) and a prospective cohort of 294 FNA biopsies (85% accuracy) from twelve independent clinical centers. This study shows that integrating high-throughput proteomics and AI technology in multi-center retrospective and prospective clinical cohorts facilitates precise disease diagnosis which is otherwise difficult to achieve by other methods.

Proteotypic Differences of Follicular-Patterned Thyroid Neoplasms.

The diagnosis of follicular-patterned thyroid tumors such as follicular thyroid adenoma (FA), follicular thyroid carcinoma (FTC), and follicular variant of papillary thyroid carcinoma (FvPTC) remains challenging. This study aimed to explore the molecular differences among these three thyroid tumors by proteomic analysis. A pressure cycling technology (PCT)-data-independent acquisition (DIA) mass spectrometry workflow was employed to investigate protein alterations in 52 formalin-fixed paraffin-embedded (FFPE) specimens: 18 FA, 15 FTC, and 19 FvPTC specimens. Immunohistochemical (IHC) analysis of 101 FA, 67 FTC, and 65 FvPTC specimens and parallel reaction monitoring (PRM) analysis of 20 FA, 20 FTC, and 20 FvPTC specimens were performed to validate protein biomarkers. A total of 4107 proteins were quantified from 52 specimens. Pairwise comparisons identified 287 differentially regulated proteins between FTC and FA, and 303 between FvPTC and FA and 88 proteins were co-dysregulated in the two comparisons. However, only 23 discriminatory proteins between FTC and FvPTC were detected. Additionally, the quantitative results for ANXA1 expression based on IHC staining and PRM-MS quantification were consistent with the proteomic results, showing that ANXA1 can be used to distinguish FvPTC from FA and FTC. The differentially regulated proteins found in this study can differentiate FA from FvPTC. In addition, ANXA1 is a promising biomarker for differentiating FvPTC from the other thyroid tumors.

Stratification of follicular thyroid tumours using data-independent acquisition proteomics and a comprehensive thyroid tissue spectral library.

Thyroid nodules occur in about 60% of the population. A major challenge in thyroid nodule diagnosis is to distinguish between follicular adenoma (FA) and carcinoma (FTC). Here, we present a comprehensive thyroid spectral library covering five types of thyroid tissues. This library includes 121 960 peptides and 9941 protein groups. This spectral library can be used to quantify up to 7863 proteins from thyroid tissues, and can also be used to develop parallel reaction monitoring (PRM) assays for targeted protein quantification. Next, to stratify follicular thyroid tumours, we compared the proteomes of 24 FA and 22 FTC samples, and identified 204 differentially expressed proteins (DEPs). Our data suggest altered ferroptosis pathways in malignant follicular carcinoma. In all, 31 selected proteins effectively distinguished follicular tumours. Of those DEPs, nine proteins were further verified by PRM in an independent cohort of 18 FA and 19 FTC. Together, we present a comprehensive spectral library for DIA and targeted proteomics analysis of thyroid tissue specimens, and identified nine proteins that could potentially distinguish FA and FTC.

Potential Use of Serum Proteomics for Monitoring COVID-19 Progression to Complement RT-PCR Detection.

RT-PCR is the primary method to diagnose COVID-19 and is also used to monitor the disease course. This approach, however, suffers from false negatives due to RNA instability and poses a high risk to medical practitioners. Here, we investigated the potential of using serum proteomics to predict viral nucleic acid positivity during COVID-19. We analyzed the proteome of 275 inactivated serum samples from 54 out of 144 COVID-19 patients and shortlisted 42 regulated proteins in the severe group and 12 in the non-severe group. Using these regulated proteins and several key clinical indexes, including days after symptoms onset, platelet counts, and magnesium, we developed two machine learning models to predict nucleic acid positivity, with an AUC of 0.94 in severe cases and 0.89 in non-severe cases, respectively. Our data suggest the potential of using a serum protein-based machine learning model to monitor COVID-19 progression, thus complementing swab RT-PCR tests. More efforts are required to promote this approach into clinical practice since mass spectrometry-based protein measurement is not currently widely accessible in clinic.

DIA-Based Proteomics Identifies IDH2 as a Targetable Regulator of Acquired Drug Resistance in Chronic Myeloid Leukemia.

Drug resistance is a critical obstacle to effective treatment in patients with chronic myeloid leukemia. To understand the underlying resistance mechanisms in response to imatinib mesylate (IMA) and adriamycin (ADR), the parental K562 cells were treated with low doses of IMA or ADR for 2 months to generate derivative cells with mild, intermediate, and severe resistance to the drugs as defined by their increasing resistance index. PulseDIA-based (DIA [data-independent acquisition]) quantitative proteomics was then employed to reveal the proteome changes in these resistant cells. In total, 7082 proteins from 98,232 peptides were identified and quantified from the dataset using four DIA software tools including OpenSWATH, Spectronaut, DIA-NN, and EncyclopeDIA. Sirtuin signaling pathway was found to be significantly enriched in both ADR-resistant and IMA-resistant K562 cells. In particular, isocitrate dehydrogenase (NADP(+)) 2 was identified as a potential drug target correlated with the drug resistance phenotype, and its inhibition by the antagonist AGI-6780 reversed the acquired resistance in K562 cells to either ADR or IMA. Together, our study has implicated isocitrate dehydrogenase (NADP(+)) 2 as a potential target that can be therapeutically leveraged to alleviate the drug resistance in K562 cells when treated with IMA and ADR.

Computational Optimization of Spectral Library Size Improves DIA-MS Proteome Coverage and Applications to 15 Tumors.

Efficient peptide and protein identifications from data-independent acquisition mass spectrometric (DIA-MS) data typically rely on a project-specific spectral library with a suitable size. Here, we describe subLib, a computational strategy for optimizing the spectral library for a specific DIA data set based on a comprehensive spectral library, requiring the preliminary analysis of the DIA data set. Compared with the pan-human library strategy, subLib achieved a 41.2% increase in peptide precursor identifications and a 35.6% increase in protein group identifications in a test data set of six colorectal tumor samples. We also applied this strategy to 389 carcinoma samples from 15 tumor data sets: up to a 39.2% increase in peptide precursor identifications and a 19.0% increase in protein group identifications were observed. Our strategy for spectral library size optimization thus successfully proved to deepen the proteome coverages of DIA-MS data.

Eleven routine clinical features predict COVID-19 severity uncovered by machine learning of longitudinal measurements.

Severity prediction of COVID-19 remains one of the major clinical challenges for the ongoing pandemic. Here, we have recruited a 144 COVID-19 patient cohort, resulting in a data matrix containing 3,065 readings for 124 types of measurements over 52 days. A machine learning model was established to predict the disease progression based on the cohort consisting of training, validation, and internal test sets. A panel of eleven routine clinical factors constructed a classifier for COVID-19 severity prediction, achieving accuracy of over 98% in the discovery set. Validation of the model in an independent cohort containing 25 patients achieved accuracy of 80%. The overall sensitivity, specificity, positive predictive value (PPV), and negative predictive value (NPV) were 0.70, 0.99, 0.93, and 0.93, respectively. Our model captured predictive dynamics of lactate dehydrogenase (LDH) and creatine kinase (CK) while their levels were in the normal range. This model is accessible at https://www.guomics.com/covidAI/ for research purpose.

The Prediction Analysis of Autistic and Schizotypal Traits in Attentional Networks.

OBJECTIVE: Empirical findings confirmed that autistic and schizotypal traits are associated with attentional function as well as include various dimensions. So far, no study has reported which dimension of these traits relates to attentional networks. This study aimed to find out whether there are associations between attentional networks and autistic traits; and between attentional networks and schizotypal traits. METHODS: A total of 449 volunteers was included in this study, and autism-spectrum quotient (AQ), schizotypal personality questionnaire (SPQ), and attention network test (ANT) were used to measure autistic traits and schizotypal traits. The three independent attentional networks, including alerting network, orienting network, and executive control network, were also measured. RESULTS: Autistic traits were associated with the orienting network, whereas schizotypal traits were associated with the orienting network and executive control network. Furthermore, attentional networks could be predicted by specific dimensions of autistic and schizotypal traits. AQ-attention switching [0.104 (-1.175- -0.025), p=0.041] and AQ-attention to detail [-0.097 (-0.798- -0.001), p=0.049] were significant predictors of orienting network and gender were significant predictor of executive network (Beta=0.107; 95% CI=-0.476-10.139; p=0.031). Whereas, schizotypal dimension "interpersonal" was a significant predictor of all three attentional networks [Alerting: 0.147 (-0.010-0.861), p=0.045; Orienting: 0.147 (0.018-0.733), p=0.040; Executive: 0.198 (0.215-1.309), p=0.006]. CONCLUSION: This study demonstrated that autistic and schizotypal traits were associated with attentional networks. The specific dimensions of autistic and schizotypal traits could predict attentional networks. Nevertheless, the attentional networks predicted with these two traits were different.

Multi-organ proteomic landscape of COVID-19 autopsies.

The molecular pathology of multi-organ injuries in COVID-19 patients remains unclear, preventing effective therapeutics development. Here, we report a proteomic analysis of 144 autopsy samples from seven organs in 19 COVID-19 patients. We quantified 11,394 proteins in these samples, in which 5,336 were perturbed in the COVID-19 patients compared to controls. Our data showed that cathepsin L1, rather than ACE2, was significantly upregulated in the lung from the COVID-19 patients. Systemic hyperinflammation and dysregulation of glucose and fatty acid metabolism were detected in multiple organs. We also observed dysregulation of key factors involved in hypoxia, angiogenesis, blood coagulation, and fibrosis in multiple organs from the COVID-19 patients. Evidence for testicular injuries includes reduced Leydig cells, suppressed cholesterol biosynthesis, and sperm mobility. In summary, this study depicts a multi-organ proteomic landscape of COVID-19 autopsies that furthers our understanding of the biological basis of COVID-19 pathology.