Dual adversarial deconfounding autoencoder for joint batch-effects removal from multi-center and multi-scanner radiomics data

Scientific Reports, Год журнала: 2023, Номер 13(1)

Опубликована: Ноя. 1, 2023

Medical imaging represents the primary tool for investigating and monitoring several diseases, including cancer. The advances in quantitative image analysis have developed towards extraction of biomarkers able to support clinical decisions. To produce robust results, multi-center studies are often set up. However, information must be denoised from confounding factors-known as batch-effect-like scanner-specific center-specific influences. Moreover, non-solid cancers, like lymphomas, effective require an imaging-based representation disease that accounts its multi-site spreading over patient's body. In this work, we address dual-factor deconfusion problem propose a algorithm harmonize patients affected by Hodgkin Lymphoma setting. We show proposed model successfully denoises data domain-specific variability (p-value < 0.001) while it coherently preserves spatial relationship between descriptions peer lesions = 0), which is strong prognostic biomarker tumor heterogeneity assessment. This harmonization step allows significantly improve performance models with respect state-of-the-art methods, enabling building exhaustive patient representations delivering more accurate analyses (p-values 0.001 training, p-values 0.05 testing). work lays groundwork performing large-scale reproducible on urgently needed convey translation into practice tools. code available GitHub at https://github.com/LaraCavinato/Dual-ADAE .

Язык: Английский

---

A Lipid Atlas of the Human Kidney

bioRxiv (Cold Spring Harbor Laboratory), Год журнала: 2022, Номер unknown

Опубликована: Апрель 7, 2022

ABSTRACT Tissue atlases provide foundational knowledge on the cellular organization and molecular distributions across classes spatial scales. Here, we construct a comprehensive spatio-molecular lipid atlas of human kidney from 29 donor tissues using integrated multimodal imaging. Our approach leverages high resolution matrix-assisted laser desorption/ionization (MALDI) imaging mass spectrometry (IMS) for untargeted mapping, stained microscopy histopathological assessment, tissue segmentation autofluorescence microscopy. With combination unsupervised, supervised, interpretive machine learning, provides multivariate profiles specific multicellular functional units (FTUs) nephron, including glomerulus, proximal tubules, thick ascending limb, distal collecting ducts. In total, consists tens thousands FTUs millions measurements. Detailed patient, clinical, histopathologic information allowed data to be mined based these features. As examples, highlight discovery how are altered with sex differences in body index.

Язык: Английский

---

Dual Adversarial Deconfounding Autoencoder for joint batch-effects removal from multi-center and multi-scanner radiomics data

bioRxiv (Cold Spring Harbor Laboratory), Год журнала: 2023, Номер unknown

Опубликована: Янв. 18, 2023

Abstract Medical imaging represents the primary tool for investigating and monitoring several diseases, including cancer. The advances in quantitative image analysis have developed towards extraction of biomarkers able to support clinical decisions. To produce robust results, multi-center studies are often set up. However, information must be denoised from confounding factors – known as batch-effect like scanner-specific center-specific influences. Moreover, non-solid cancers, lymphomas, effective require an imaging-based representation disease that accounts its multi-site spreading over patient’s body. In this work, we address dual-factor deconfusion problem propose a algorithm harmonize patients affected by Hodgkin Lymphoma setting. We show proposed model successfully denoises data domain-specific variability while it coherently preserves spatial relationship between descriptions peer lesions, which is strong prognostic biomarker tumor heterogeneity assessment. This harmonization step allows significantly improve performance models, enabling building exhaustive patient representations delivering more accurate analyses. work lays groundwork performing large-scale reproducible analyses on urgently needed convey translation into practice tools. code available GitHub at link

Язык: Английский

---

Disentangling plant response to biotic and abiotic stress using HIVE, a novel tool to perform unpaired multi-transcriptomics integration

bioRxiv (Cold Spring Harbor Laboratory), Год журнала: 2024, Номер unknown

Опубликована: Март 5, 2024

Abstract All organisms are subjected to multiple stresses usually occurring at the same time, requiring activation of appropriate signalling pathways respond all or by prioritizing response one stress factor. Plants, as sessile organisms, particularly impacted constantly changing environment that is often unfavourable even hostile. Because experimental complexity studying organism stressors simultaneously, experiments conducted considering individual factor time. An alternative consists in performing silico integration those data on single response. Currently used methods integrate unpaired consist meta-analysis finding differentially expressed genes for each condition separately and then selecting commonly regulated ones. Although these approaches allowed find valuable results, they mainly identify specific signatures very few signature responding lack modulated differently condition. For this purpose, we developed HIVE (Horizontal Integration analysis using Variational AutoEncoders) single-stress transcriptomics datasets composed experiments. Briefly, coupled a variational autoencoder, alleviates batch effects, with random forest regression SHAP explainer select relevant specifically stresses. We illustrate functionality study transcriptional changes several different plants namely Arabidopsis thaliana , rice, maize, wheat, grapevine peanut collecting publicly available stress, either biotic and/or abiotic, jointly analyse them. performed better than differential expression analysis, state-of-the-art tool horizontal allowing novel promising candidates responsible triggering effective defence responses

Язык: Английский

---

DNA methylation classifier to diagnose pancreatic ductal adenocarcinoma metastases from different anatomical sites

Clinical Epigenetics, Год журнала: 2024, Номер 16(1)

Опубликована: Ноя. 10, 2024

Abstract Background We have recently constructed a DNA methylation classifier that can discriminate between pancreatic ductal adenocarcinoma (PAAD) liver metastasis and intrahepatic cholangiocarcinoma (iCCA) with high accuracy ( PAAD-iCCA-Classifier ). PAAD is one of the leading causes cancer unknown primary diagnosis based on exclusion other malignancies. Therefore, our focus was to investigate whether be used diagnose metastases from sites. Methods For this scope, anomaly detection filter initial expanded by 8 additional mimicker carcinomas, amounting total 10 carcinomas in negative class. validated updated version validation set, which consisted biological cohort n = 3579) technical 15). then assessed performance test included positive control 16 various sites 124 samples consisting 96 breast 18 anatomical 28 carcinoma brain. Results The achieved 98.21% samples, ones it reached 100%. also correctly identified 15/16 (93.75%) as PAAD, control, classified 122/124 (98.39%) for 97.85% overall set. dataset explore organotropism observed are distinct peritoneal carcinomatosis characterized specific copy number alterations hypomethylation enhancers involved epithelial-mesenchymal-transition. Conclusions (available at https://classifier.tgc-research.de/ ) accurately classify metastatic serve diagnostic aid.

Язык: Английский

---

Computational Methods for Data Integration and Imputation of Missing Values in Omics Datasets

PROTEOMICS, Год журнала: 2024, Номер 25(1-2)

Опубликована: Дек. 30, 2024

ABSTRACT Molecular profiling of different omic ‐modalities (e.g., DNA methylomics, transcriptomics, proteomics) in biological systems represents the basis for research and clinical decision‐making. Measurement‐specific biases, so‐called batch effects, often hinder integration independently acquired datasets, missing values further hamper applicability typical data processing algorithms. In addition to careful experimental design, well‐defined standards acquisition exchange, alleviation these phenomena particularly requires a dedicated preprocessing pipeline. This review aims give comprehensive overview computational methods value imputation analyses. We provide formal definitions mechanisms propose novel statistical taxonomy especially presence data. Based on an automated document search systematic literature review, we describe 32 distinct from five main methodological categories, as well 37 algorithms separate categories. Additionally, this highlights multiple quantitative evaluation aid researchers selecting suitable set their work. Finally, work provides integrated discussion relevance effects omics with corresponding method recommendations. then three‐step workflow study conception final analysis deduce perspectives future research. Eventually, present flow chart exemplary decision trees practitioners selection specific approaches studies.

Язык: Английский

---

Epistemic uncertainty challenges aging clock reliability in predicting rejuvenation effects

bioRxiv (Cold Spring Harbor Laboratory), Год журнала: 2023, Номер unknown

Опубликована: Дек. 4, 2023

Abstract Epigenetic aging clocks have been widely used to validate rejuvenation effects during cellular reprogramming. However, these predictions are unverifiable because the true biological age of reprogrammed cells remains unknown. We present an analytical framework consider from uncertainty perspective. Our analysis reveals that DNA methylation profiles across reprogramming poorly represented in data train clock models, thus introducing high epistemic estimations. Moreover, different published inconsistent, with some even suggesting zero or negative rejuvenation. While not questioning possibility reversal, we show challenges reliability observed vitro before pluripotency and throughout embryogenesis. Conversely, our method a significant increase after vivo recommend including estimation future models avoid risk misinterpreting results prediction.

Язык: Английский

---