Dimitrios Mathios
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Research Subjects
Dimitrios Mathios has a total of 148 co-authors. Here are the top co-authors by number of shared publications:
- Jillian Phallen
The Sidney Kimmel Comprehensive Cancer Center, Johns Hopkins University School of Medicine, Baltimore, Maryland
24 shared publications - Akshaya V. Annapragada
The Sidney Kimmel Comprehensive Cancer Center, Johns Hopkins University School of Medicine, Baltimore, MD, USA
23 shared publications - Daniel C. Bruhm
Sidney Kimmel Comprehensive Cancer Center, Johns Hopkins University School of Medicine, Baltimore, MD 21287, USA
23 shared publications - Jamie E. Medina
The Sidney Kimmel Comprehensive Cancer Center, Johns Hopkins University School of Medicine, Baltimore, Maryland
23 shared publications - Noushin Niknafs
Sidney Kimmel Comprehensive Cancer Center, Johns Hopkins University School of Medicine, Baltimore, MD 21287, USA
23 shared publications - Robert B. Scharpf
The Sidney Kimmel Comprehensive Cancer Center, Johns Hopkins University School of Medicine, Baltimore, Maryland
23 shared publications
Latest Publications
- Open Access
- 13 April 2026
- Published by Oxford University Press (OUP) in Neuro-Oncology
- Vol. 28 (7) , 1649-1661
Laser interstitial thermal therapy enhances bidirectional blood-brain barrier permeability in glioblastoma
Background: Laser interstitial thermal therapy (LITT) is a minimally invasive treatment for glioblastoma that increases blood-brain barrier (BBB) permeability. However, the mechanisms and spatiotemporal features of this effect remain unclear. It is also unknown whether LITT promotes release of circulating tumor DNA (ctDNA). Methods: Using our previously developed LITT mouse model, we employed single-cell RNA-sequencing (scRNA-seq) to investigate gene expression changes in endothelial cells following LITT in the naive mouse brain. Brains were also harvested at multiple time points to assess LITT effects on tight junction (TJ) integrity and transcytosis via immunofluorescence and transmission electron microscopy. Human glioblastoma tissues were analyzed to monitor ambient caveolae-mediated transcytosis in tumor regions. Finally, ctDNA levels in mouse plasma were quantified using digital-droplet PCR following LITT. Results: LITT triggered downregulation of pathways related to cell-cell junctions and upregulation of transcytosis. LITT transiently disrupted TJ integrity for 7 days up to 100 µm away from the ablation zone. Further, LITT increased caveolae-mediated transcytosis in endothelial cells for 21 days, peaking at 14 days, in part due to suppression of the transcytosis inhibitor Mfsd2a. LITT facilitated ctDNA detection in mouse plasma at 1 h, 3 days, 7 days, and 14 days post-procedure. Conclusions: Our study demonstrates that LITT transiently disrupts the BBB in the periphery of the ablation with endothelial TJ disruption and transcytosis upregulation following distinct time courses. LITT effects on the BBB are bidirectional and enhance ctDNA release in addition to enabling brain entry of agents.
- 13 April 2026
- Published by Oxford University Press (OUP) in Neuro-Oncology
- Vol. 28 (7) , 1649-1661
Open AccessLaser interstitial thermal therapy enhances bidirectional blood-brain barrier permeability in glioblastoma
Background: Laser interstitial thermal therapy (LITT) is a minimally invasive treatment for glioblastoma that increases blood-brain barrier (BBB) permeability. However, the mechanisms and spatiotemporal features of this effect remain unclear. It is also unknown whether LITT promotes release of circulating tumor DNA (ctDNA). Methods: Using our previously developed LITT mouse model, we employed single-cell RNA-sequencing (scRNA-seq) to investigate gene expression changes in endothelial cells following LITT in the naive mouse brain. Brains were also harvested at multiple time points to assess LITT effects on tight junction (TJ) integrity and transcytosis via immunofluorescence and transmission electron microscopy. Human glioblastoma tissues were analyzed to monitor ambient caveolae-mediated transcytosis in tumor regions. Finally, ctDNA levels in mouse plasma were quantified using digital-droplet PCR following LITT. Results: LITT triggered downregulation of pathways related to cell-cell junctions and upregulation of transcytosis. LITT transiently disrupted TJ integrity for 7 days up to 100 µm away from the ablation zone. Further, LITT increased caveolae-mediated transcytosis in endothelial cells for 21 days, peaking at 14 days, in part due to suppression of the transcytosis inhibitor Mfsd2a. LITT facilitated ctDNA detection in mouse plasma at 1 h, 3 days, 7 days, and 14 days post-procedure. Conclusions: Our study demonstrates that LITT transiently disrupts the BBB in the periphery of the ablation with endothelial TJ disruption and transcytosis upregulation following distinct time courses. LITT effects on the BBB are bidirectional and enhance ctDNA release in addition to enabling brain entry of agents.
- 4 March 2026
- Published by American Association for the Advancement of Science (AAAS) in Science Translational Medicine
- Vol. 18 (839) , eadw2603
Cell-free DNA fragmentomes for noninvasive detection of liver cirrhosis and other diseases
Accessible liquid biopsies, including analyses of genome-wide cell-free DNA (cfDNA) fragmentation, are emerging for early detection of cancer but remain largely unexplored in other diseases. Here, we used whole-genome sequencing to examine cfDNA fragmentomes in 1576 individuals, including those with liver disease or with other morbidities such as vascular, autoimmune, and neurodegenerative conditions. As a prototype for disease-specific cfDNA fragmentomic biomarkers, we developed a machine learning classifier that detected early liver disease, advanced fibrosis, and cirrhosis with high sensitivity in separate discovery ( n = 423) and validation cohorts ( n = 221) and had limited cross-reactivity for other diseases. Genome-wide fragmentome and methylome analyses revealed liver-derived and immune-mediated changes in cfDNA in the circulation of individuals affected with liver disease. Fragmentomic changes were also observed across a range of other human morbidities and reflected disease-specific changes in the circulation. A machine learning model using cfDNA fragmentomes predicted overall survival in separate morbidity discovery ( n = 571) and validation cohorts ( n = 231). These analyses demonstrate the connection between cfDNA fragmentomes and an individual’s physiologic state and provide previously unrecognized possibilities for cfDNA liquid biopsies across human disease.
- 4 March 2026
- Published by American Association for the Advancement of Science (AAAS) in Science Translational Medicine
- Vol. 18 (839) , eadw2603
Cell-free DNA fragmentomes for noninvasive detection of liver cirrhosis and other diseases
Accessible liquid biopsies, including analyses of genome-wide cell-free DNA (cfDNA) fragmentation, are emerging for early detection of cancer but remain largely unexplored in other diseases. Here, we used whole-genome sequencing to examine cfDNA fragmentomes in 1576 individuals, including those with liver disease or with other morbidities such as vascular, autoimmune, and neurodegenerative conditions. As a prototype for disease-specific cfDNA fragmentomic biomarkers, we developed a machine learning classifier that detected early liver disease, advanced fibrosis, and cirrhosis with high sensitivity in separate discovery ( n = 423) and validation cohorts ( n = 221) and had limited cross-reactivity for other diseases. Genome-wide fragmentome and methylome analyses revealed liver-derived and immune-mediated changes in cfDNA in the circulation of individuals affected with liver disease. Fragmentomic changes were also observed across a range of other human morbidities and reflected disease-specific changes in the circulation. A machine learning model using cfDNA fragmentomes predicted overall survival in separate morbidity discovery ( n = 571) and validation cohorts ( n = 231). These analyses demonstrate the connection between cfDNA fragmentomes and an individual’s physiologic state and provide previously unrecognized possibilities for cfDNA liquid biopsies across human disease.
- Open Access
- 11 December 2025
- Published by American Association for Cancer Research (AACR)
Supplementary Tables from Detecting Liver Cancer Using Cell-Free DNA Fragmentomes
Supplementary Table S1 shows Clinical information of US/EU patients analyzed. Supplementary Table S2 is a summary of whole genome cfDNA analyses. Supplementary Table S3 shows clinical information of Hong Kong patients analyzed. Supplementary Table S4 shows the lowest scoring transcription factors in US/EU cohort samples. Supplementary Table S5 lists the plasma volume and cfDNA amounts used in genomic libraries.
- 11 December 2025
- Published by American Association for Cancer Research (AACR)
Open AccessSupplementary Tables from Detecting Liver Cancer Using Cell-Free DNA Fragmentomes
Supplementary Table S1 shows Clinical information of US/EU patients analyzed. Supplementary Table S2 is a summary of whole genome cfDNA analyses. Supplementary Table S3 shows clinical information of Hong Kong patients analyzed. Supplementary Table S4 shows the lowest scoring transcription factors in US/EU cohort samples. Supplementary Table S5 lists the plasma volume and cfDNA amounts used in genomic libraries.
- Open Access
- 11 December 2025
- Published by American Association for Cancer Research (AACR)
Supplementary Figures S1-S14 from Detecting Liver Cancer Using Cell-Free DNA Fragmentomes
Supplementary Figure S1 shows contributions of individual genomic regions to the final trained DELFI surveillance model. Supplementary Figure S2 shows DELFI scores in individuals without cancer by age and sex. Supplementary Figure S3 shows DELFI scores in high-risk individuals without cancer across racial or ethnic groups. Supplementary Figure S4 shows DELFI scores in cirrhotic patients in relation to severity. Supplementary Figure S5 shows DELFI scores in individuals in relation to BMI. Supplementary Figure S6 Shows Performance of alternative DELFI models. Supplementary Figure S7 shows Correlation between rank ordered DELFI. Supplementary Figure S8 shows DELFI scores in patients with HCC were related to tumor characteristics. Supplementary Figure S9 shows DELFI scores in by etiology of underlying liver disease. Supplementary Figure S10 shows the DELFI score in patients with HCC is correlated to AFP levels. Supplementary Figure S11 shows the correlation of fragmentation profiles. Supplementary Figure S12 shows Genome wide fragmentation profiles in the Hong Kong validation cohort Supplementary Figure S13 shows chromosomal copy number in Hong Kong cohort. Supplementary Figure S14 shows the implementation of DELFI in liver cancer screening.
- 11 December 2025
- Published by American Association for Cancer Research (AACR)
Open AccessSupplementary Figures S1-S14 from Detecting Liver Cancer Using Cell-Free DNA Fragmentomes
Supplementary Figure S1 shows contributions of individual genomic regions to the final trained DELFI surveillance model. Supplementary Figure S2 shows DELFI scores in individuals without cancer by age and sex. Supplementary Figure S3 shows DELFI scores in high-risk individuals without cancer across racial or ethnic groups. Supplementary Figure S4 shows DELFI scores in cirrhotic patients in relation to severity. Supplementary Figure S5 shows DELFI scores in individuals in relation to BMI. Supplementary Figure S6 Shows Performance of alternative DELFI models. Supplementary Figure S7 shows Correlation between rank ordered DELFI. Supplementary Figure S8 shows DELFI scores in patients with HCC were related to tumor characteristics. Supplementary Figure S9 shows DELFI scores in by etiology of underlying liver disease. Supplementary Figure S10 shows the DELFI score in patients with HCC is correlated to AFP levels. Supplementary Figure S11 shows the correlation of fragmentation profiles. Supplementary Figure S12 shows Genome wide fragmentation profiles in the Hong Kong validation cohort Supplementary Figure S13 shows chromosomal copy number in Hong Kong cohort. Supplementary Figure S14 shows the implementation of DELFI in liver cancer screening.
- Open Access
- 4 August 2025
- Published by American Association for Cancer Research (AACR)
Figure 3 from Detection of Brain Cancer Using Genome-wide Cell-free DNA Fragmentomes
Survival analyses of patients with molecularly defined untreated GBM using ARTEMIS–DELFI model components. A, The horizontal axis indicates the classifier score cutoff used to dichotomize the GBM patient cohort to low- and high-score categories. The vertical axis indicates the P value of the univariate survival analysis of the Cox proportional hazard ratios for each ARTEMIS–DELFI submodel at all evaluated score cutoffs. Each submodel of the ARTEMIS–DELFI model is represented with a different color. Kaplan–Meier survival analyses of the GBM cohort for the chromosomal arm model (B) and the epigenetic model (C) both at a score cutoff of 0.75. P values are calculated based on the log-rank test, and the estimated HR and 95% CI are shown. TE, transposable element, SINE, short interspersed nuclear elements, LINE, long interspersed nuclear elements.
- 4 August 2025
- Published by American Association for Cancer Research (AACR)
Open AccessFigure 3 from Detection of Brain Cancer Using Genome-wide Cell-free DNA Fragmentomes
Survival analyses of patients with molecularly defined untreated GBM using ARTEMIS–DELFI model components. A, The horizontal axis indicates the classifier score cutoff used to dichotomize the GBM patient cohort to low- and high-score categories. The vertical axis indicates the P value of the univariate survival analysis of the Cox proportional hazard ratios for each ARTEMIS–DELFI submodel at all evaluated score cutoffs. Each submodel of the ARTEMIS–DELFI model is represented with a different color. Kaplan–Meier survival analyses of the GBM cohort for the chromosomal arm model (B) and the epigenetic model (C) both at a score cutoff of 0.75. P values are calculated based on the log-rank test, and the estimated HR and 95% CI are shown. TE, transposable element, SINE, short interspersed nuclear elements, LINE, long interspersed nuclear elements.
- Open Access
- 4 August 2025
- Published by American Association for Cancer Research (AACR)
Table 1 from Detection of Brain Cancer Using Genome-wide Cell-free DNA Fragmentomes
Patient characteristics for the discovery and validation cohorts.
- 4 August 2025
- Published by American Association for Cancer Research (AACR)
Open AccessTable 1 from Detection of Brain Cancer Using Genome-wide Cell-free DNA Fragmentomes
Patient characteristics for the discovery and validation cohorts.
- Open Access
- 4 August 2025
- Published by American Association for Cancer Research (AACR)
Supplementary Tables S1-S7 from Detection of Brain Cancer Using Genome-wide Cell-free DNA Fragmentomes
Supplementary Table S1. Clinical information, sequencing statistics, and ARTEMIS-DELFI scores for cfDNA analyses in Discovery cohort. Supplementary Table S2. Clinical information, sequencing statistics, and ARTEMIS-DELFI scores for cfDNA analyses in Validation cohort. Supplementary Table S3. Summary of mutations detected by plasma-based targeted sequencing panel. Supplementary Table S4. Correlation of TF expression between GBM and other tumor types. Supplementary Table S5. Summary of complete blood count for patients with gliomas or non-neoplastic neurological conditions. Supplementary Table S6. Summary of ChIP-Seq peaks used in DECIFER analysis. Supplementary Table S7. Correlation of TFBS relative coverage differences with expression differences between tumor/cell types and whole blood.
- 4 August 2025
- Published by American Association for Cancer Research (AACR)
Open AccessSupplementary Tables S1-S7 from Detection of Brain Cancer Using Genome-wide Cell-free DNA Fragmentomes
Supplementary Table S1. Clinical information, sequencing statistics, and ARTEMIS-DELFI scores for cfDNA analyses in Discovery cohort. Supplementary Table S2. Clinical information, sequencing statistics, and ARTEMIS-DELFI scores for cfDNA analyses in Validation cohort. Supplementary Table S3. Summary of mutations detected by plasma-based targeted sequencing panel. Supplementary Table S4. Correlation of TF expression between GBM and other tumor types. Supplementary Table S5. Summary of complete blood count for patients with gliomas or non-neoplastic neurological conditions. Supplementary Table S6. Summary of ChIP-Seq peaks used in DECIFER analysis. Supplementary Table S7. Correlation of TFBS relative coverage differences with expression differences between tumor/cell types and whole blood.
- Open Access
- 4 August 2025
- Published by American Association for Cancer Research (AACR)
Supplementary Figures S1-S10 from Detection of Brain Cancer Using Genome-wide Cell-free DNA Fragmentomes
Supplementary Figure S1. Schematic representation of the Ensemble model architecture used for brain cancer prediction. Supplementary Figure S2. Feature importance of the ARTEMIS-DELFI classifier by feature family. Supplementary Figure S3. ARTEMIS-DELFI score distribution across genomic library batch preparation. Supplementary Figure S4. Correlation of brain tumor size with ARTEMIS-DELFI scores. Supplementary Figure S5. ARTEMIS-DELFI scores correlate with Ki-67 proliferation indices in tumor samples. Supplementary Figure S6. Fragment length cumulative distributions of cfDNA in brain tumor patients by mutation type. Supplementary Figure S7. Schematic representation of the DECIFER methodology. Supplementary Figure S8. Heatmap representation of TF RNA expression levels by tissue type. Supplementary Figure S9. DECIFER correlation plots from the Discovery and the Validation cohorts with different tumor and tissue types. Supplementary Figure S10. Modelling the potential implementation of ARTEMIS-DELFI in the workup of headaches for brain tumors.
- 4 August 2025
- Published by American Association for Cancer Research (AACR)
Open AccessSupplementary Figures S1-S10 from Detection of Brain Cancer Using Genome-wide Cell-free DNA Fragmentomes
Supplementary Figure S1. Schematic representation of the Ensemble model architecture used for brain cancer prediction. Supplementary Figure S2. Feature importance of the ARTEMIS-DELFI classifier by feature family. Supplementary Figure S3. ARTEMIS-DELFI score distribution across genomic library batch preparation. Supplementary Figure S4. Correlation of brain tumor size with ARTEMIS-DELFI scores. Supplementary Figure S5. ARTEMIS-DELFI scores correlate with Ki-67 proliferation indices in tumor samples. Supplementary Figure S6. Fragment length cumulative distributions of cfDNA in brain tumor patients by mutation type. Supplementary Figure S7. Schematic representation of the DECIFER methodology. Supplementary Figure S8. Heatmap representation of TF RNA expression levels by tissue type. Supplementary Figure S9. DECIFER correlation plots from the Discovery and the Validation cohorts with different tumor and tissue types. Supplementary Figure S10. Modelling the potential implementation of ARTEMIS-DELFI in the workup of headaches for brain tumors.
- Open Access
- 4 August 2025
- Published by American Association for Cancer Research (AACR)
Figure 4 from Detection of Brain Cancer Using Genome-wide Cell-free DNA Fragmentomes
CfDNA fragmentation and mutations in plasma of patients with HGGs. A, The table shows the number and percentage of patients analyzed and detected with the ARTEMIS–DELFI assay, the tumor-guided mutational panel approach, or both. B, Histogram representation of the sensitivity of HGG detection by the mutation panel or the ARTEMIS–DELFI approach. The score threshold corresponding to 90% specificity is applied to determine the samples detected by ARTEMIS–DELFI approach. The analytical specificity of the tumor-informed mutation panel assay is 99.9%. C, Fragment length distribution of tumor-specific mutant fragments vs. fragments with no mutations, germline mutations or known clonal hematopoiesis of indeterminate potential (CHIP) mutations. Tumor-specific mutations are residing in shorter fragments than wild-type (WT) or nontumor mutant fragments (median length of 146 bp vs. 167 bp, Wilcoxon P < 0.0001).
- 4 August 2025
- Published by American Association for Cancer Research (AACR)
Open AccessFigure 4 from Detection of Brain Cancer Using Genome-wide Cell-free DNA Fragmentomes
CfDNA fragmentation and mutations in plasma of patients with HGGs. A, The table shows the number and percentage of patients analyzed and detected with the ARTEMIS–DELFI assay, the tumor-guided mutational panel approach, or both. B, Histogram representation of the sensitivity of HGG detection by the mutation panel or the ARTEMIS–DELFI approach. The score threshold corresponding to 90% specificity is applied to determine the samples detected by ARTEMIS–DELFI approach. The analytical specificity of the tumor-informed mutation panel assay is 99.9%. C, Fragment length distribution of tumor-specific mutant fragments vs. fragments with no mutations, germline mutations or known clonal hematopoiesis of indeterminate potential (CHIP) mutations. Tumor-specific mutations are residing in shorter fragments than wild-type (WT) or nontumor mutant fragments (median length of 146 bp vs. 167 bp, Wilcoxon P < 0.0001).
- Open Access
- 4 August 2025
- Published by American Association for Cancer Research (AACR)
Figure 2 from Detection of Brain Cancer Using Genome-wide Cell-free DNA Fragmentomes
ARTEMIS–DELFI performance for detection of brain cancers. ARTEMIS–DELFI score distribution across noncancer individuals and patients with brain tumor, stratified by grade and histology, in the discovery cohort. A, The boxplot shows the median ARTEMIS–DELFI score and the IQR with the individual sample values overlaid as dots. The noncancer neurologic controls, although overall exhibiting low ARTEMIS–DELFI scores, have a higher score than the screening population group (median score of 0.25 vs. 0.11; P < 0.0001). The noncancer cases with or without neurologic conditions have a lower ARTEMIS–DELFI score compared with brain tumor cases, and there is an incremental increase in the ARTEMIS–DELFI score between the LGGs (median score of 0.46) and HGGs (median score of 0.71). Within primary brain tumors, the highest median DELFI score is observed in GBMs (median score 0.72). Metastatic brain tumors (including lung, breast, ovarian, melanoma, and kidney cancers) have a high score (median of 0.86). The center line in the boxplots represents the median, the top limit of the boxplots represents the third quantile (75th percentile), the bottom limit of the boxplots represents the first quantile (25th percentile), the top whiskers is the maximum value of the data that are within 1.5 times the IQR over the 75th percentile, and the bottom whisker is the minimum value of the data that are within 1.5 times the IQR under the 25th percentile. Rec, recurrent. B, ROC analyses of the overall discovery cohort as well as by grade. The faded gray vertical lines in the ROC figures represents a 95% and 90% specificity as decision boundaries. C, The boxplot shows the median ARTEMIS–DELFI score and the IQR with the individual sample values overlaid as dots in the validation cohort. The noncancer neurologic controls exhibit a low ARTEMIS–DELFI score (median of 0.11), with gliomas (median of 0.70) and brain metastases (median of 0.86) showing elevated scores similar to our discovery cohort. D, Analysis of an ARTEMIS–DELFI fixed model and score cutoff of 0.30 determined from the discovery cohort was applied in the validation cohort. The performance of this classifier in the independent cohort was similar to the discovery cohort in both specificity (left) and sensitivity (right). The number of samples in the discovery and validation sets are indicated in the labels of the horizontal axis. The intervals presented reflect a 90% CI.
- 4 August 2025
- Published by American Association for Cancer Research (AACR)
Open AccessFigure 2 from Detection of Brain Cancer Using Genome-wide Cell-free DNA Fragmentomes
ARTEMIS–DELFI performance for detection of brain cancers. ARTEMIS–DELFI score distribution across noncancer individuals and patients with brain tumor, stratified by grade and histology, in the discovery cohort. A, The boxplot shows the median ARTEMIS–DELFI score and the IQR with the individual sample values overlaid as dots. The noncancer neurologic controls, although overall exhibiting low ARTEMIS–DELFI scores, have a higher score than the screening population group (median score of 0.25 vs. 0.11; P < 0.0001). The noncancer cases with or without neurologic conditions have a lower ARTEMIS–DELFI score compared with brain tumor cases, and there is an incremental increase in the ARTEMIS–DELFI score between the LGGs (median score of 0.46) and HGGs (median score of 0.71). Within primary brain tumors, the highest median DELFI score is observed in GBMs (median score 0.72). Metastatic brain tumors (including lung, breast, ovarian, melanoma, and kidney cancers) have a high score (median of 0.86). The center line in the boxplots represents the median, the top limit of the boxplots represents the third quantile (75th percentile), the bottom limit of the boxplots represents the first quantile (25th percentile), the top whiskers is the maximum value of the data that are within 1.5 times the IQR over the 75th percentile, and the bottom whisker is the minimum value of the data that are within 1.5 times the IQR under the 25th percentile. Rec, recurrent. B, ROC analyses of the overall discovery cohort as well as by grade. The faded gray vertical lines in the ROC figures represents a 95% and 90% specificity as decision boundaries. C, The boxplot shows the median ARTEMIS–DELFI score and the IQR with the individual sample values overlaid as dots in the validation cohort. The noncancer neurologic controls exhibit a low ARTEMIS–DELFI score (median of 0.11), with gliomas (median of 0.70) and brain metastases (median of 0.86) showing elevated scores similar to our discovery cohort. D, Analysis of an ARTEMIS–DELFI fixed model and score cutoff of 0.30 determined from the discovery cohort was applied in the validation cohort. The performance of this classifier in the independent cohort was similar to the discovery cohort in both specificity (left) and sensitivity (right). The number of samples in the discovery and validation sets are indicated in the labels of the horizontal axis. The intervals presented reflect a 90% CI.