Sophia Apostolidou
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Previous Organizations / Institutes
The organizations / institutes from the scholar's other articles in the past ten years
Research Subjects
Sophia Apostolidou has a total of 179 co-authors. Here are the top co-authors by number of shared publications:
- Aleksandra Gentry-Maharaj
Department of Women’s Cancer, Institute for Women’s Health, University College London, London, UK
26 shared publications - Usha Menon
University College London, London, United Kingdom
25 shared publications - Andy Ryan
MRC Clinical Trials Unit, Institute of Clinical Trials and Methodology, University College London, London, UK
16 shared publications - Jane M. Lange
Cancer Early Detection Advanced Research Knight Cancer Institute Oregon Health Sciences University Portland Oregon USA
13 shared publications - Jason Xu
Department of Statistical Science, Duke University
13 shared publications - Kemal Caglar Gogebakan
Fred Hutch Cancer Center 1 Seattle, WA
13 shared publications
Latest Publications
- Open Access
- 11 November 2026
- Published by Springer Nature in Nature
- p. 1-10
Evolutionary dynamics in the years preceding acute myeloid leukaemia
Somatic evolution in ageing tissues underlies many cancers 1,2 . However, our quantitative understanding of the rules governing this pre-cancerous evolution remains incomplete. Here we exploit a unique collection of serial blood samples 3 , collected annually for up to a decade and beginning as early as 15 years before diagnosis of acute myeloid leukaemia, to provide a quantitative description of pre-cancerous evolutionary dynamics. Using deep duplex sequencing and evolutionary theory, we quantify the acquisition ages and fitness effects of the key driver events in acute myeloid leukaemia development. The first driver mutations are typically acquired in the first few decades of life, when the blood remains highly polyclonal. These early slow-growing clones subsequently acquire multiple further driver mutations that confer selective advantages up to 100-fold larger than the early drivers. These faster-growing clones harbouring multiple driver mutations can cause complete somatic sweeps of the blood decades before diagnosis, a feature strongly associated with future acute myeloid leukaemia. Our data show that the clonal dynamics of blood are consistent with a set of remarkably simple evolutionary rules that strike a balance between chance and determinism.
- 11 November 2026
- Published by Springer Nature in Nature
- p. 1-10
Open AccessEvolutionary dynamics in the years preceding acute myeloid leukaemia
Somatic evolution in ageing tissues underlies many cancers 1,2 . However, our quantitative understanding of the rules governing this pre-cancerous evolution remains incomplete. Here we exploit a unique collection of serial blood samples 3 , collected annually for up to a decade and beginning as early as 15 years before diagnosis of acute myeloid leukaemia, to provide a quantitative description of pre-cancerous evolutionary dynamics. Using deep duplex sequencing and evolutionary theory, we quantify the acquisition ages and fitness effects of the key driver events in acute myeloid leukaemia development. The first driver mutations are typically acquired in the first few decades of life, when the blood remains highly polyclonal. These early slow-growing clones subsequently acquire multiple further driver mutations that confer selective advantages up to 100-fold larger than the early drivers. These faster-growing clones harbouring multiple driver mutations can cause complete somatic sweeps of the blood decades before diagnosis, a feature strongly associated with future acute myeloid leukaemia. Our data show that the clonal dynamics of blood are consistent with a set of remarkably simple evolutionary rules that strike a balance between chance and determinism.
- Open Access
- 9 October 2026
- Published by MDPI AG in Diagnostics
- Vol. 16 (20) , 3283
Graph Neural Network Modelling of Pre-Diagnostic Serum Proteomics for Pancreatic Cancer Detection
Background/Objectives: Pancreatic ductal adenocarcinoma (PDAC) has exceptionally high mortality, largely because of late diagnosis, while established blood biomarkers such as CA19-9 have limited sensitivity for early detection. We evaluated whether modelling serum proteomic measurements as supervised pairwise statistical graphs using synolitic graph neural networks (SGNNs) could provide a useful framework for pre-diagnostic PDAC classification. Methods: We analysed 217 pre-diagnostic serum samples from UKCTOCS, comprising 100 PDAC samples from 75 women and 117 control samples, profiled for 97 proteins. Samples were divided into a development set collected 0–1 year before diagnosis (n = 110) and a temporal holdout set collected 1–2 years before diagnosis (n = 107); 25 PDAC participants contributed samples to both windows. The final SGNN used fold-internal mutual-information selection of 30 proteins, pairwise RBF-SVM-derived edge information, minimum-connected sparsification, and a 10-member GATv2 ensemble per cross-validation fold. Conventional machine-learning comparators were tuned using development data only. Results: The SGNN achieved a mean 5-fold development ROC-AUC of 75.33 ± 10.43%, lower than the tuned Random Forest (81.67%); the other tuned comparators achieved 77.67% (XGBoost), 77.17% (logistic regression), and 73.17% (SVM). Averaging predictions across the five-fold-specific SGNN ensembles yielded a temporal-holdout ROC-AUC of 64.35% (95% CI 52.6–74.6%). In the participant-independent subset of 25 PDAC cases not represented in development and 57 controls, ROC-AUC was 66.04% (95% CI 52.56–78.88%). Conclusions: SGNNs provide a feasible graph-based representation of pre-diagnostic proteomic data but did not outperform optimised conventional machine-learning methods within development cross-validation. The temporal and participant-independent performance estimates warrant further investigation in larger fully independent cohorts. A graph-specific predictive advantage was not directly tested—no such advantage was demonstrated in this dataset—and the clinical utility has not been established.
- 9 October 2026
- Published by MDPI AG in Diagnostics
- Vol. 16 (20) , 3283
Open AccessGraph Neural Network Modelling of Pre-Diagnostic Serum Proteomics for Pancreatic Cancer Detection
Background/Objectives: Pancreatic ductal adenocarcinoma (PDAC) has exceptionally high mortality, largely because of late diagnosis, while established blood biomarkers such as CA19-9 have limited sensitivity for early detection. We evaluated whether modelling serum proteomic measurements as supervised pairwise statistical graphs using synolitic graph neural networks (SGNNs) could provide a useful framework for pre-diagnostic PDAC classification. Methods: We analysed 217 pre-diagnostic serum samples from UKCTOCS, comprising 100 PDAC samples from 75 women and 117 control samples, profiled for 97 proteins. Samples were divided into a development set collected 0–1 year before diagnosis (n = 110) and a temporal holdout set collected 1–2 years before diagnosis (n = 107); 25 PDAC participants contributed samples to both windows. The final SGNN used fold-internal mutual-information selection of 30 proteins, pairwise RBF-SVM-derived edge information, minimum-connected sparsification, and a 10-member GATv2 ensemble per cross-validation fold. Conventional machine-learning comparators were tuned using development data only. Results: The SGNN achieved a mean 5-fold development ROC-AUC of 75.33 ± 10.43%, lower than the tuned Random Forest (81.67%); the other tuned comparators achieved 77.67% (XGBoost), 77.17% (logistic regression), and 73.17% (SVM). Averaging predictions across the five-fold-specific SGNN ensembles yielded a temporal-holdout ROC-AUC of 64.35% (95% CI 52.6–74.6%). In the participant-independent subset of 25 PDAC cases not represented in development and 57 controls, ROC-AUC was 66.04% (95% CI 52.56–78.88%). Conclusions: SGNNs provide a feasible graph-based representation of pre-diagnostic proteomic data but did not outperform optimised conventional machine-learning methods within development cross-validation. The temporal and participant-independent performance estimates warrant further investigation in larger fully independent cohorts. A graph-specific predictive advantage was not directly tested—no such advantage was demonstrated in this dataset—and the clinical utility has not been established.
- Open Access
- 30 April 2026
- Published by Elsevier in SSRN Electronic Journal
Immunoaffinity‑Enriched Extracellular Vesicle Proteomics for High‑Grade Serous Ovarian Cancer - A UKOPS Nested Case‑Control Analysis
Background: Extracellular vesicles (EVs) represent a promising source of blood-based biomarkers for ovarian cancer detection, with growing interest in their application for early diagnosis. A systematic evaluation of EV subpopulations derived from plasma and serum is needed to determine their comparability for immunoaffinity-enriched proteomic profiling and to enable broader use of clinical biobank collections.Methods: We performed immunoaffinity-enriched proteomic profiling of EV from 243 paired EDTA plasma and serum samples within a nested case–control cohort from the UK Ovarian Cancer Population Study, including controls, benign, borderline, and high-grade serous ovarian cancer (HGSOC) cases. Integrative analyses assessed proteomic depth, plasma–serum concordance, and classification performance using machine learning approaches. Key findings were evaluated across EV subpopulations to determine their suitability for biomarker discovery and application across diverse clinical sample types.Finding: Plasma-derived EV subpopulations showed greater proteomic depth than serum, with 6439 versus 5451 protein groups identified and 2838 versus 840 proteins detected at 50% prevalence. Despite this difference, protein intensities were highly concordant between matrices (r=0.86, p<0.001), and this concordance was maintained across disease groups and cancer stages. For distinguishing HGSOC from non‑cancers, the EV‑augmented CA‑125 model achieved superior performance, with an AUROC of 0.97 in plasma and 0.96 in serum. It improved plasma sensitivity to 99% (95% CI 89–100) and specificity to 85% (95% CI 68–91), compared with 96% (95% CI 82–100) sensitivity and 65% (95% CI 51–73) specificity for CA‑125 alone, while maintaining similarly high discrimination in serum.Interpretation: Surface epitope-targeted EV subpopulation enrichment yields biologically specific proteomic profiling of HGSOC with highly concordant proteomic biomarker signatures across plasma and serum, supporting their interchangeable use.
- 30 April 2026
- Published by Elsevier in SSRN Electronic Journal
Open AccessImmunoaffinity‑Enriched Extracellular Vesicle Proteomics for High‑Grade Serous Ovarian Cancer - A UKOPS Nested Case‑Control Analysis
Background: Extracellular vesicles (EVs) represent a promising source of blood-based biomarkers for ovarian cancer detection, with growing interest in their application for early diagnosis. A systematic evaluation of EV subpopulations derived from plasma and serum is needed to determine their comparability for immunoaffinity-enriched proteomic profiling and to enable broader use of clinical biobank collections.Methods: We performed immunoaffinity-enriched proteomic profiling of EV from 243 paired EDTA plasma and serum samples within a nested case–control cohort from the UK Ovarian Cancer Population Study, including controls, benign, borderline, and high-grade serous ovarian cancer (HGSOC) cases. Integrative analyses assessed proteomic depth, plasma–serum concordance, and classification performance using machine learning approaches. Key findings were evaluated across EV subpopulations to determine their suitability for biomarker discovery and application across diverse clinical sample types.Finding: Plasma-derived EV subpopulations showed greater proteomic depth than serum, with 6439 versus 5451 protein groups identified and 2838 versus 840 proteins detected at 50% prevalence. Despite this difference, protein intensities were highly concordant between matrices (r=0.86, p<0.001), and this concordance was maintained across disease groups and cancer stages. For distinguishing HGSOC from non‑cancers, the EV‑augmented CA‑125 model achieved superior performance, with an AUROC of 0.97 in plasma and 0.96 in serum. It improved plasma sensitivity to 99% (95% CI 89–100) and specificity to 85% (95% CI 68–91), compared with 96% (95% CI 82–100) sensitivity and 65% (95% CI 51–73) specificity for CA‑125 alone, while maintaining similarly high discrimination in serum.Interpretation: Surface epitope-targeted EV subpopulation enrichment yields biologically specific proteomic profiling of HGSOC with highly concordant proteomic biomarker signatures across plasma and serum, supporting their interchangeable use.
- Open Access
- 3 November 2025
- Published by American Association for Cancer Research (AACR)
Table S3 from Estimating the Opportunity for Early Detection of Ovarian Cancer Using Individual-Patient Data from a Large Randomized Controlled Trial
Table S3 shows natural history estimates from a model with MMS episode-specific sensitivity.
- 3 November 2025
- Published by American Association for Cancer Research (AACR)
Open AccessTable S3 from Estimating the Opportunity for Early Detection of Ovarian Cancer Using Individual-Patient Data from a Large Randomized Controlled Trial
Table S3 shows natural history estimates from a model with MMS episode-specific sensitivity.
- Open Access
- 3 November 2025
- Published by American Association for Cancer Research (AACR)
Figure S2 from Estimating the Opportunity for Early Detection of Ovarian Cancer Using Individual-Patient Data from a Large Randomized Controlled Trial
Figure S2 shows posterior predictive checks for the primary analysis across all cancers and trial arms, comparing observed data with model predictions for clinical and screen-detected cancers.
- 3 November 2025
- Published by American Association for Cancer Research (AACR)
Open AccessFigure S2 from Estimating the Opportunity for Early Detection of Ovarian Cancer Using Individual-Patient Data from a Large Randomized Controlled Trial
Figure S2 shows posterior predictive checks for the primary analysis across all cancers and trial arms, comparing observed data with model predictions for clinical and screen-detected cancers.
- Open Access
- 3 November 2025
- Published by American Association for Cancer Research (AACR)
Supplementary Methods from Estimating the Opportunity for Early Detection of Ovarian Cancer Using Individual-Patient Data from a Large Randomized Controlled Trial
Supplementary Methods describe the model selection process using approximate leave-one-out cross-validation, posterior predictive checks comparing predicted and observed cancer incidence, and the estimation of empirical screening sensitivities from the fitted natural history model.
- 3 November 2025
- Published by American Association for Cancer Research (AACR)
Open AccessSupplementary Methods from Estimating the Opportunity for Early Detection of Ovarian Cancer Using Individual-Patient Data from a Large Randomized Controlled Trial
Supplementary Methods describe the model selection process using approximate leave-one-out cross-validation, posterior predictive checks comparing predicted and observed cancer incidence, and the estimation of empirical screening sensitivities from the fitted natural history model.
- Open Access
- 3 November 2025
- Published by American Association for Cancer Research (AACR)
Figure S4 from Estimating the Opportunity for Early Detection of Ovarian Cancer Using Individual-Patient Data from a Large Randomized Controlled Trial
Figure S4. Posterior predictive checks for the primary analysis (all three trial arms) among non-HGSC cancers.
- 3 November 2025
- Published by American Association for Cancer Research (AACR)
Open AccessFigure S4 from Estimating the Opportunity for Early Detection of Ovarian Cancer Using Individual-Patient Data from a Large Randomized Controlled Trial
Figure S4. Posterior predictive checks for the primary analysis (all three trial arms) among non-HGSC cancers.
- Open Access
- 3 November 2025
- Published by American Association for Cancer Research (AACR)
Figure S1 from Estimating the Opportunity for Early Detection of Ovarian Cancer Using Individual-Patient Data from a Large Randomized Controlled Trial
Figure S1 shows predictive model fits from approximate leave-one-out cross-validation for the full dataset and for cancer cases only, highlighting model selection steps and the final chosen model.
- 3 November 2025
- Published by American Association for Cancer Research (AACR)
Open AccessFigure S1 from Estimating the Opportunity for Early Detection of Ovarian Cancer Using Individual-Patient Data from a Large Randomized Controlled Trial
Figure S1 shows predictive model fits from approximate leave-one-out cross-validation for the full dataset and for cancer cases only, highlighting model selection steps and the final chosen model.
- Open Access
- 3 November 2025
- Published by American Association for Cancer Research (AACR)
Table S2 from Estimating the Opportunity for Early Detection of Ovarian Cancer Using Individual-Patient Data from a Large Randomized Controlled Trial
Table S2 shows predictive fit and preclinical detectable phase estimates for the best-fitting models, alongside key parameter estimates.
- 3 November 2025
- Published by American Association for Cancer Research (AACR)
Open AccessTable S2 from Estimating the Opportunity for Early Detection of Ovarian Cancer Using Individual-Patient Data from a Large Randomized Controlled Trial
Table S2 shows predictive fit and preclinical detectable phase estimates for the best-fitting models, alongside key parameter estimates.
- Open Access
- 3 November 2025
- Published by American Association for Cancer Research (AACR)
Figure S6 from Estimating the Opportunity for Early Detection of Ovarian Cancer Using Individual-Patient Data from a Large Randomized Controlled Trial
Figure S6 shows the sensitivity of parameter estimates to model structure, displaying medians and credible intervals for onset rate, mean preclinical phase, and MMS/USS sensitivities across well-fitting models.
- 3 November 2025
- Published by American Association for Cancer Research (AACR)
Open AccessFigure S6 from Estimating the Opportunity for Early Detection of Ovarian Cancer Using Individual-Patient Data from a Large Randomized Controlled Trial
Figure S6 shows the sensitivity of parameter estimates to model structure, displaying medians and credible intervals for onset rate, mean preclinical phase, and MMS/USS sensitivities across well-fitting models.