Subhajit Dutta
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Research Subjects
Subhajit Dutta has a total of 33 co-authors. Here are the top co-authors by number of shared publications:
Matiar Rahman
Department of Biochemistry & Molecular Biology, University of Rajshahi, Rajshahi, Bangladesh
2 shared publicationsMohammad Sayful Islam
Department of Pharmacy, Mawlana Bhashani Science and Technology University, Tangail, Bangladesh
2 shared publications- Moyen Uddin Pk
Department of Quality Control & Assurance Jist Life Care Ltd Dhaka Bangladesh
2 shared publications Rabiul Islam Talukder
Department of Clinical Biochemistry, Popular Diagnostic Centre Ltd., Dhaka, Bangladesh
2 shared publicationsRumana Pervin
Department of Biochemistry & Molecular Biology, University of Rajshahi, Rajshahi, Bangladesh
2 shared publicationsAbiodun Samuel Oyedele
1 shared publication
Latest Publications
- Open Access
- 27 October 2026
- Published by Frontiers Media SA in Frontiers in Pharmacology
- Vol. 17, 1983363
The mitochondrion as a mechanometabolic conductor of cancer metastasis: pharmacological insights from the mitochondria–peroxisome interface
Metastasis demands more than sustained ATP production: cancer cells must move organelles, remodel membranes, manage oxidative stress, and reprogram metabolism as they pass from invasion to distant colonisation. Mitochondria sit at the centre of this adaptation. This Review organises their contribution around four linked axes: deployment to the cell cortex and invadopodia; reactive oxygen species and metabolite signalling; VDAC1-dependent metabolite, calcium, apoptotic and mtDNA gating; and HIF-1α–BNIP3-dependent remodelling of the mitochondrial pool. We then ask how this mitochondrial programme is altered by peroxisomes. The two organelles share components of the fission machinery, coordinate lipid-derived substrate use and cellular redox balance, and divide tasks in hydrogen-peroxide control, ether-lipid synthesis, organelle quality control and MAVS signalling. These connections matter pharmacologically. A DRP1- or redox-directed intervention may affect both organelles, whereas VDAC1 provides a more mitochondria-restricted target logic. Peroxisomal pathways are also context dependent: fatty-acid shortening and ether-lipid synthesis can increase metabolic flexibility and membrane remodelling, yet the same pathways may raise oxidative stress or create ferroptotic susceptibility. By separating established molecular links from mechanisms that remain untested in metastatic models, we propose a hierarchical view in which mitochondria remain the principal mechanometabolic conductor, while peroxisomes set important boundary conditions for substrate use, redox signalling, immune responses and therapeutic selectivity. We also state the limitation this framing carries. The mitochondrial arm of the framework rests on direct experimental evidence, whereas a causal contribution of mitochondria–peroxisome contacts to invasion, transit or colonisation has not been demonstrated: it is inferred here from contact biology characterised without testing metastatic endpoints, from division machinery shared between the two organelles, and from biochemical coupling. The interface is therefore advanced as a testable framework for target selection and pharmacological attribution, not as an established driver of metastasis.
- 27 October 2026
- Published by Frontiers Media SA in Frontiers in Pharmacology
- Vol. 17, 1983363
Open AccessThe mitochondrion as a mechanometabolic conductor of cancer metastasis: pharmacological insights from the mitochondria–peroxisome interface
Metastasis demands more than sustained ATP production: cancer cells must move organelles, remodel membranes, manage oxidative stress, and reprogram metabolism as they pass from invasion to distant colonisation. Mitochondria sit at the centre of this adaptation. This Review organises their contribution around four linked axes: deployment to the cell cortex and invadopodia; reactive oxygen species and metabolite signalling; VDAC1-dependent metabolite, calcium, apoptotic and mtDNA gating; and HIF-1α–BNIP3-dependent remodelling of the mitochondrial pool. We then ask how this mitochondrial programme is altered by peroxisomes. The two organelles share components of the fission machinery, coordinate lipid-derived substrate use and cellular redox balance, and divide tasks in hydrogen-peroxide control, ether-lipid synthesis, organelle quality control and MAVS signalling. These connections matter pharmacologically. A DRP1- or redox-directed intervention may affect both organelles, whereas VDAC1 provides a more mitochondria-restricted target logic. Peroxisomal pathways are also context dependent: fatty-acid shortening and ether-lipid synthesis can increase metabolic flexibility and membrane remodelling, yet the same pathways may raise oxidative stress or create ferroptotic susceptibility. By separating established molecular links from mechanisms that remain untested in metastatic models, we propose a hierarchical view in which mitochondria remain the principal mechanometabolic conductor, while peroxisomes set important boundary conditions for substrate use, redox signalling, immune responses and therapeutic selectivity. We also state the limitation this framing carries. The mitochondrial arm of the framework rests on direct experimental evidence, whereas a causal contribution of mitochondria–peroxisome contacts to invasion, transit or colonisation has not been demonstrated: it is inferred here from contact biology characterised without testing metastatic endpoints, from division machinery shared between the two organelles, and from biochemical coupling. The interface is therefore advanced as a testable framework for target selection and pharmacological attribution, not as an established driver of metastasis.
- Open Access
- 9 September 2026
- Published by Frontiers Media SA in Frontiers in Systems Biology
- Vol. 6, 1859256
Covariate-adjusted AP-1 motif architecture and footprinting track olaparib-adaptive chromatin remodelling
Transcription-factor involvement in drug-adaptive chromatin remodelling is often inferred from motif enrichment in differential ATAC-seq peaks, but such analyses can be distorted by sequence composition, peak length, baseline accessibility and redundant motif models. We reanalysed 105,048 consensus ATAC-seq peaks from an olaparib-adapted Kuramochi ovarian-cancer continuum using HOMER-calibrated AP-1/bZIP models with coordinate-level de-duplication. At T320, 27,154 peaks gained and 30,050 lost accessibility. AP-1 motif-occurrence density remained higher in gained chromatin under complementary adjustments: an all-peak spline-adjusted Poisson model gave a rate ratio of 1.90 (95% CI 1.80–2.01), overlap weighting gave 1.95 (1.82–2.11), and 14,046 covariate-matched pairs gave 1.97 (1.90–2.05). Following 35,325 fixed motif-positive/motif-negative peak pairs across the adapted series showed a positive accessibility trend of 0.0363 contrast units per dose doubling (95% CI 0.0195–0.0531). Raw paired-end ATAC-seq footprinting provided an orthogonal signal: AP-1 motif intervals scored above same-length motif-free controls selected within the same accessible peaks in every sample, and the AP-1-specific footprint score increased by 0.00509 per dose doubling after replicate adjustment (HC3 95% CI 0.00081–0.00937; exact stratified permutation P = 0.00211). In a matched-background genome-wide scan, related AP-1/bZIP models occupied the nine highest ranks. Within gained chromatin, motif-positive peaks showed greater adjusted overlap with pooled public AP-1 summits (OR 3.82, 95% CI 3.52–4.15); after collapsing related files by study, 71 study clusters gave a random-effects OR of 2.68 (2.51–2.87), with substantial heterogeneity and a prediction interval of 1.58–4.54. Multi-model single-cell and independent bulk-RNA analyses indicated context-dependent downstream AP-1 output, whereas the two valid comparable drug-free CRISPR contrasts were not FDR-significant. Exploratory nearest-TSS mapping of AP-1-rich gained peaks identified 899 protein-coding genes; recurring GO, Reactome and KEGG themes and STRING associations were used only to provide qualitative, hypothesis-generating functional context. Together, the data support a robust association between AP-1-compatible chromatin architecture and olaparib adaptation, while not establishing AP-1 occupancy, necessity or sufficiency in the studied cells.
- 9 September 2026
- Published by Frontiers Media SA in Frontiers in Systems Biology
- Vol. 6, 1859256
Open AccessCovariate-adjusted AP-1 motif architecture and footprinting track olaparib-adaptive chromatin remodelling
Transcription-factor involvement in drug-adaptive chromatin remodelling is often inferred from motif enrichment in differential ATAC-seq peaks, but such analyses can be distorted by sequence composition, peak length, baseline accessibility and redundant motif models. We reanalysed 105,048 consensus ATAC-seq peaks from an olaparib-adapted Kuramochi ovarian-cancer continuum using HOMER-calibrated AP-1/bZIP models with coordinate-level de-duplication. At T320, 27,154 peaks gained and 30,050 lost accessibility. AP-1 motif-occurrence density remained higher in gained chromatin under complementary adjustments: an all-peak spline-adjusted Poisson model gave a rate ratio of 1.90 (95% CI 1.80–2.01), overlap weighting gave 1.95 (1.82–2.11), and 14,046 covariate-matched pairs gave 1.97 (1.90–2.05). Following 35,325 fixed motif-positive/motif-negative peak pairs across the adapted series showed a positive accessibility trend of 0.0363 contrast units per dose doubling (95% CI 0.0195–0.0531). Raw paired-end ATAC-seq footprinting provided an orthogonal signal: AP-1 motif intervals scored above same-length motif-free controls selected within the same accessible peaks in every sample, and the AP-1-specific footprint score increased by 0.00509 per dose doubling after replicate adjustment (HC3 95% CI 0.00081–0.00937; exact stratified permutation P = 0.00211). In a matched-background genome-wide scan, related AP-1/bZIP models occupied the nine highest ranks. Within gained chromatin, motif-positive peaks showed greater adjusted overlap with pooled public AP-1 summits (OR 3.82, 95% CI 3.52–4.15); after collapsing related files by study, 71 study clusters gave a random-effects OR of 2.68 (2.51–2.87), with substantial heterogeneity and a prediction interval of 1.58–4.54. Multi-model single-cell and independent bulk-RNA analyses indicated context-dependent downstream AP-1 output, whereas the two valid comparable drug-free CRISPR contrasts were not FDR-significant. Exploratory nearest-TSS mapping of AP-1-rich gained peaks identified 899 protein-coding genes; recurring GO, Reactome and KEGG themes and STRING associations were used only to provide qualitative, hypothesis-generating functional context. Together, the data support a robust association between AP-1-compatible chromatin architecture and olaparib adaptation, while not establishing AP-1 occupancy, necessity or sufficiency in the studied cells.
- Open Access
- 3 September 2026
- Published by Frontiers Media SA in Frontiers in Cell and Developmental Biology
- Vol. 14, 1930678
Mechanical observables of cancer invasion: actin, adhesion, and metastatic plasticity
Cancer invasion is commonly classified through epithelial-mesenchymal plasticity, yet cells with similar transcriptional states can differ substantially in force production, cytoskeletal organization and collective mobility. We critically examine whether mechanical measurements can supply information that is not already contained in molecular or histopathological classifications. The available evidence does not justify a universal five-dimensional phase space. It supports a more restrained hierarchy: primary measurements (filament orientation, cortical tension, substrate traction, cell and nuclear boundaries, and velocity fields), derived descriptors (polarity, field-specific normalized nematic order, shape and density metrics), and tissue-scale constructions (orientation fields, defects and inferred unjamming states). These levels are coupled and often mathematically dependent. We correct several common conflations. Traction-force microscopy recovers substrate traction through an inverse mechanical model; it does not directly measure intrinsic cortical active stress. Isotropic cortical contractility must be separated from anisotropic nematic stress. Structural actin order must likewise be distinguished from apolar alignment calculated from an optical-flow field; the latter is a dynamic image-derived descriptor, not a direct measurement of filament orientation or active stress. The vertex model threshold p 0 * ≃ 3.81 is a model parameter, not a universal cutoff for segmented tumor cells. Patient-tissue unjamming maps instead use multivariate shape-density information, including cell-and-nucleus shape (CeNuS) and nuclear number density. We evaluate evidence supporting and opposing the independence of unjamming from epithelial-mesenchymal transition, the context-dependent effects of contractility, and the proposed causal role of topological defects. Our conclusion is deliberately conditional: mechanical observables are most credible as complementary, scale-specific measurements whose incremental value must be established against EMT/EMP state, tumor grade and tissue architecture in the same specimens. This formulation retains the explanatory power of mechanics while making its assumptions, dependencies and clinical limits explicit.
- 3 September 2026
- Published by Frontiers Media SA in Frontiers in Cell and Developmental Biology
- Vol. 14, 1930678
Open AccessMechanical observables of cancer invasion: actin, adhesion, and metastatic plasticity
Cancer invasion is commonly classified through epithelial-mesenchymal plasticity, yet cells with similar transcriptional states can differ substantially in force production, cytoskeletal organization and collective mobility. We critically examine whether mechanical measurements can supply information that is not already contained in molecular or histopathological classifications. The available evidence does not justify a universal five-dimensional phase space. It supports a more restrained hierarchy: primary measurements (filament orientation, cortical tension, substrate traction, cell and nuclear boundaries, and velocity fields), derived descriptors (polarity, field-specific normalized nematic order, shape and density metrics), and tissue-scale constructions (orientation fields, defects and inferred unjamming states). These levels are coupled and often mathematically dependent. We correct several common conflations. Traction-force microscopy recovers substrate traction through an inverse mechanical model; it does not directly measure intrinsic cortical active stress. Isotropic cortical contractility must be separated from anisotropic nematic stress. Structural actin order must likewise be distinguished from apolar alignment calculated from an optical-flow field; the latter is a dynamic image-derived descriptor, not a direct measurement of filament orientation or active stress. The vertex model threshold p 0 * ≃ 3.81 is a model parameter, not a universal cutoff for segmented tumor cells. Patient-tissue unjamming maps instead use multivariate shape-density information, including cell-and-nucleus shape (CeNuS) and nuclear number density. We evaluate evidence supporting and opposing the independence of unjamming from epithelial-mesenchymal transition, the context-dependent effects of contractility, and the proposed causal role of topological defects. Our conclusion is deliberately conditional: mechanical observables are most credible as complementary, scale-specific measurements whose incremental value must be established against EMT/EMP state, tumor grade and tissue architecture in the same specimens. This formulation retains the explanatory power of mechanics while making its assumptions, dependencies and clinical limits explicit.
- Open Access
A replicated patient-specific component of tumour telomere length across two pan-cancer cohorts
Bulk telomere length measured from tumour sequencing is routinely read as a property of the cancer cells. A tumour specimen is a mixture, however, and the patient who supplies it has a telomere length of their own. Here I re-analyse published pan-cancer telomere estimates and ask how much of a tumour’s telomere length is the patient’s. A calibration step comes first. Whole-genome and low-pass estimates recover the known cross-sectional attrition of leukocyte telomeres with age (−26.6 bp per year in blood normals), whereas whole-exome estimates do not: adjusted for cancer type, sequencing centre and sex the exome slope is −0.6 bp per year, and in 684 blood-normal aliquots sequenced by both assays the whole-genome estimate declines at −38.9 bp per year while the exome estimate from the same DNA does not decline at all (difference −41.5 bp per year, P = 3 × 10⁻¹⁰). Because exome data are 78.6% of the resource, downstream analyses use only the calibrated libraries. Within those, tumour telomere length tracks the patient’s matched normal (Spearman ρ = +0.395; 22 of 23 cancer types positive), and this replicates in a second consortium measured with a different estimator (PCAWG ρ = +0.472, 24 of 24 histologies). Entering cancer type, sequencing centre and library type as fixed effects leaves β = 0.385, and the coefficient is unmoved by age, sex, purity, leukocyte fraction, ploidy, coverage and continental ancestry (range 0.406–0.429). Pure normal-cell admixture is rejected as the sole explanation: the mixture model requires β_H = 1 and β_H×p = −1, and both are rejected jointly ( P = 0.001). Tumour purity, leukocyte fraction and age each explain only ∼1–3% of within-cohort variance, and none alters the cross-cancer ranking. The between-cohort coefficient is not interpretable: it appears to show one-to-one correspondence with tissue-of-origin telomere length, but that value depends on which tissue supplies the reference and on the spread of the predictor, and falls to 0.44 with an organ-matched solid-tissue normal. Bulk tumour telomere length is therefore a composite phenotype containing a replicated patient-specific component. Telomere biomarker studies should carry matched normal telomere length as a covariate rather than treating tumour telomere length as tumour-intrinsic.
Open AccessA replicated patient-specific component of tumour telomere length across two pan-cancer cohorts
Bulk telomere length measured from tumour sequencing is routinely read as a property of the cancer cells. A tumour specimen is a mixture, however, and the patient who supplies it has a telomere length of their own. Here I re-analyse published pan-cancer telomere estimates and ask how much of a tumour’s telomere length is the patient’s. A calibration step comes first. Whole-genome and low-pass estimates recover the known cross-sectional attrition of leukocyte telomeres with age (−26.6 bp per year in blood normals), whereas whole-exome estimates do not: adjusted for cancer type, sequencing centre and sex the exome slope is −0.6 bp per year, and in 684 blood-normal aliquots sequenced by both assays the whole-genome estimate declines at −38.9 bp per year while the exome estimate from the same DNA does not decline at all (difference −41.5 bp per year, P = 3 × 10⁻¹⁰). Because exome data are 78.6% of the resource, downstream analyses use only the calibrated libraries. Within those, tumour telomere length tracks the patient’s matched normal (Spearman ρ = +0.395; 22 of 23 cancer types positive), and this replicates in a second consortium measured with a different estimator (PCAWG ρ = +0.472, 24 of 24 histologies). Entering cancer type, sequencing centre and library type as fixed effects leaves β = 0.385, and the coefficient is unmoved by age, sex, purity, leukocyte fraction, ploidy, coverage and continental ancestry (range 0.406–0.429). Pure normal-cell admixture is rejected as the sole explanation: the mixture model requires β_H = 1 and β_H×p = −1, and both are rejected jointly ( P = 0.001). Tumour purity, leukocyte fraction and age each explain only ∼1–3% of within-cohort variance, and none alters the cross-cancer ranking. The between-cohort coefficient is not interpretable: it appears to show one-to-one correspondence with tissue-of-origin telomere length, but that value depends on which tissue supplies the reference and on the spread of the predictor, and falls to 0.44 with an organ-matched solid-tissue normal. Bulk tumour telomere length is therefore a composite phenotype containing a replicated patient-specific component. Telomere biomarker studies should carry matched normal telomere length as a covariate rather than treating tumour telomere length as tumour-intrinsic.
- 6 August 2026
- Published by Springer Nature in Nature Reviews Chemistry
- Vol. 10 (9) , 605
Missing marks made visible
- 6 August 2026
- Published by Springer Nature in Nature Reviews Chemistry
- Vol. 10 (9) , 605
Missing marks made visible
- Open Access
Temporal order separates generic persistence from cue-directed commitment in cell migration
A cell track contains not only a set of movements but also an order. We asked whether that order reveals behaviour that speed, endpoint displacement, and conventional persistence measures miss. For each trajectory, we compared the recorded sequence with an exact reference in which the same displacement steps were rearranged; the observed-minus-reference difference quantifies persistence created by step order. In 48,134 MCF10DCIS.com trajectories, most of a reproducible MYO10–collagen interaction in directional persistence disappeared when temporal order was removed. In 131 single MCF10A haptotaxis tracks, widening the fibronectin pattern did not produce a resolved global increase in order; it shifted short- timescale organization from the gradient axis toward lateral movement, with an intrinsic lateral-order difference of 0.2176 (95% hierarchical interval 0.1274–0.3171). In MDA-MB-231 cells migrating toward EGF, PFKL-N702T produced fewer up-gradient steps and shorter up-gradient runs while the mean-speed difference remained unresolved. These directional effects remained after comparisons balanced for measured speed and duration, whereas three cue-independent serial-order measures showed no reduction larger than a locked 0.10-SD margin at the track level. Across independent endothelial, epithelial, and cancer-cell archives, the same calculation distinguished sustained, short-lived, and sign-reversing regimes; a stationary angular hidden-state model failed all eight frozen long-lag targets. Temporal order therefore separates three biological questions: how much a cell moves, whether it continues moving persistently, and whether that persistence remains aligned with an environmental cue. We provide a documented workflow that calculates these readouts from standard x–y tracking tables. Impact statement: The order of otherwise similar migration steps reveals whether cells merely continue moving or sustain commitment to an environmental cue.
Open AccessTemporal order separates generic persistence from cue-directed commitment in cell migration
A cell track contains not only a set of movements but also an order. We asked whether that order reveals behaviour that speed, endpoint displacement, and conventional persistence measures miss. For each trajectory, we compared the recorded sequence with an exact reference in which the same displacement steps were rearranged; the observed-minus-reference difference quantifies persistence created by step order. In 48,134 MCF10DCIS.com trajectories, most of a reproducible MYO10–collagen interaction in directional persistence disappeared when temporal order was removed. In 131 single MCF10A haptotaxis tracks, widening the fibronectin pattern did not produce a resolved global increase in order; it shifted short- timescale organization from the gradient axis toward lateral movement, with an intrinsic lateral-order difference of 0.2176 (95% hierarchical interval 0.1274–0.3171). In MDA-MB-231 cells migrating toward EGF, PFKL-N702T produced fewer up-gradient steps and shorter up-gradient runs while the mean-speed difference remained unresolved. These directional effects remained after comparisons balanced for measured speed and duration, whereas three cue-independent serial-order measures showed no reduction larger than a locked 0.10-SD margin at the track level. Across independent endothelial, epithelial, and cancer-cell archives, the same calculation distinguished sustained, short-lived, and sign-reversing regimes; a stationary angular hidden-state model failed all eight frozen long-lag targets. Temporal order therefore separates three biological questions: how much a cell moves, whether it continues moving persistently, and whether that persistence remains aligned with an environmental cue. We provide a documented workflow that calculates these readouts from standard x–y tracking tables. Impact statement: The order of otherwise similar migration steps reveals whether cells merely continue moving or sustain commitment to an environmental cue.
- Open Access
Transcriptome-inferred CIN-like fields mark antigen-presentation-low, immune-cold spatial neighborhoods
Cancer immune escape is usually interpreted as either a tumor-cell-intrinsic antigen-presentation defect or an immune-cold microenvironment. How these states are spatially arranged relative to chromosomal-instability-like genotype states remains unresolved. Here, spatial transcriptomic, single-cell and clinical transcriptomic analyses show that transcriptome-inferred CIN-like fields mark antigen-presentation-low, immune-cold neighborhoods. Across 14 melanoma tissue sections, the interaction between inferred copy-number-like burden and chromosome-expression deviation was associated with reduced MHC-I antigen-presentation activity after adjustment for tumor state, immune inflammation, technical covariates and spatial coordinates. CIN-high/AP-low spots formed coherent neighborhoods enriched for AP-low neighbors but depleted of immune-high and IFN-high neighbors under block-preserving spatial nulls. Two breast cancer Visium sections reproduced AP-low neighborhood topology, while malignant melanoma single-cell RNA-seq supported a tumor-intrinsic CNA-by-chromosome-deviation association with AP repression. Patient-level TCGA analyses linked the adverse genotype-ecotype score to poor melanoma survival. These results define transcriptome-inferred CIN-like fields as spatial correlates of immune-cold antigen-presentation failure.
Open AccessTranscriptome-inferred CIN-like fields mark antigen-presentation-low, immune-cold spatial neighborhoods
Cancer immune escape is usually interpreted as either a tumor-cell-intrinsic antigen-presentation defect or an immune-cold microenvironment. How these states are spatially arranged relative to chromosomal-instability-like genotype states remains unresolved. Here, spatial transcriptomic, single-cell and clinical transcriptomic analyses show that transcriptome-inferred CIN-like fields mark antigen-presentation-low, immune-cold neighborhoods. Across 14 melanoma tissue sections, the interaction between inferred copy-number-like burden and chromosome-expression deviation was associated with reduced MHC-I antigen-presentation activity after adjustment for tumor state, immune inflammation, technical covariates and spatial coordinates. CIN-high/AP-low spots formed coherent neighborhoods enriched for AP-low neighbors but depleted of immune-high and IFN-high neighbors under block-preserving spatial nulls. Two breast cancer Visium sections reproduced AP-low neighborhood topology, while malignant melanoma single-cell RNA-seq supported a tumor-intrinsic CNA-by-chromosome-deviation association with AP repression. Patient-level TCGA analyses linked the adverse genotype-ecotype score to poor melanoma survival. These results define transcriptome-inferred CIN-like fields as spatial correlates of immune-cold antigen-presentation failure.
- Open Access
A pan-cancer regulatory atlas of 6,983 GWAS variants prioritizes recurrent regulatory annotations and candidate programs at cancer risk loci
Genome-wide association studies have identified thousands of cancer risk variants in non-coding regions, yet their regulatory mechanisms remain largely uncharacterized. Here we present a regulatory annotation atlas of 6,983 genome-wide significant variants across 23 cancer types, scored using multimodal AlphaGenome predictions and integrated with ENCODE-4, Roadmap Epigenomics, and JASPAR 2024 annotations. Most variants (70.5%) fall outside annotated cis-regulatory elements; 27.7% overlap enhancers and 1.4% overlap promoters. Comparison with 6,626 position-matched eQTL control variants suggests that enhancer-classified variants carry 1.86-fold higher predicted effects (P = 10 − ⁹⁴) and promoter variants 7.84-fold (P = 2.5 × 10 − ¹⁹). A composite prioritization score (RegVar-basic, excluding GWAS-derived pleiotropy and TF disruption, AUC = 0.650; RegVar-full, AUC = 0.675) outperforms CADD (0.499) and LINSIGHT (0.558) in this cancer-gene discrimination benchmark. Within-locus ranking across 2,626 GTEx DAP-G eQTL credible sets shows that RegVar identifies the highest-posterior-probability variant in 47.3% of loci (P = 7.0 × 10 − ¹³), while CADD performs at chance. Predicted target genes show 67.7% concordance with GTEx eQTL assignments. Permutation-controlled motif analysis highlights NFKB1/NF-κB, STAT1, IRF1, and ARNT as exploratory permutation-enriched candidate transcription factors at cancer risk loci. This atlas provides a resource for interpreting non-coding cancer susceptibility variants. Because AlphaGenome uses expression-related training data, GTEx-based validations should be interpreted as partially orthogonal rather than fully independent.
Open AccessA pan-cancer regulatory atlas of 6,983 GWAS variants prioritizes recurrent regulatory annotations and candidate programs at cancer risk loci
Genome-wide association studies have identified thousands of cancer risk variants in non-coding regions, yet their regulatory mechanisms remain largely uncharacterized. Here we present a regulatory annotation atlas of 6,983 genome-wide significant variants across 23 cancer types, scored using multimodal AlphaGenome predictions and integrated with ENCODE-4, Roadmap Epigenomics, and JASPAR 2024 annotations. Most variants (70.5%) fall outside annotated cis-regulatory elements; 27.7% overlap enhancers and 1.4% overlap promoters. Comparison with 6,626 position-matched eQTL control variants suggests that enhancer-classified variants carry 1.86-fold higher predicted effects (P = 10 − ⁹⁴) and promoter variants 7.84-fold (P = 2.5 × 10 − ¹⁹). A composite prioritization score (RegVar-basic, excluding GWAS-derived pleiotropy and TF disruption, AUC = 0.650; RegVar-full, AUC = 0.675) outperforms CADD (0.499) and LINSIGHT (0.558) in this cancer-gene discrimination benchmark. Within-locus ranking across 2,626 GTEx DAP-G eQTL credible sets shows that RegVar identifies the highest-posterior-probability variant in 47.3% of loci (P = 7.0 × 10 − ¹³), while CADD performs at chance. Predicted target genes show 67.7% concordance with GTEx eQTL assignments. Permutation-controlled motif analysis highlights NFKB1/NF-κB, STAT1, IRF1, and ARNT as exploratory permutation-enriched candidate transcription factors at cancer risk loci. This atlas provides a resource for interpreting non-coding cancer susceptibility variants. Because AlphaGenome uses expression-related training data, GTEx-based validations should be interpreted as partially orthogonal rather than fully independent.
- 11 November 2025
- Published by Springer Nature in Nature Reviews Endocrinology
- Vol. 22 (3) , 137
CD38 as a driver of ageing
- 11 November 2025
- Published by Springer Nature in Nature Reviews Endocrinology
- Vol. 22 (3) , 137
CD38 as a driver of ageing
- Open Access
- 30 September 2025
- Published by eLife Sciences Publications, Ltd in eLife
- Vol. 14
Telomeres control human telomerase (TERT) expression through non-telomeric TRF2
The function of the human telomerase reverse transcriptase (referred hereafter as TERT) in the synthesis and maintenance of chromosome ends, or telomeres, is widely understood. Whether and how telomeres, on the other hand, influence TERT regulation is relatively less studied. We found TERT was transcriptionally altered depending on telomere length (TL). This resulted from TL-dependent binding of TRF2 between telomeres and the TERT promoter. TERT promoter-bound TRF2 was non-telomeric and did not involve the looping of telomeres to the TERT promoter. Cell lines from different tissue types fibrosarcoma (HT1080), colon cancer (HCT116), and breast cancer (MDA-MB-231), engineered for either telomere elongation/shortening, gave an increase/decrease in TERT, respectively. Mechanistically, we show TERT promoter-bound non-telomeric TRF2 recruits the canonical PRC2-complex, inducing repressor histone H3K27-trimethylation in a TL-dependent fashion. This was further supported by TL-dependent promoter activity from an exogenously inserted TERT reporter. Increase in TL over days followed by a gradual decline, resulted in activation followed by repression of TERT in a concerted manner, further implicating TL as a key factor for TERT regulation. Notably, on reprogramming primary fibroblasts to induced pluripotent stem cells (iPSCs), TRF2 loss from the TERT promoter was evident along with telomere elongation and TERT upregulation. Conversely, on telomere shortening in iPSCs, TERT promoter-bound TRF2 was restored with a marked reduction in TERT, further supporting the causal role of TL in TERT transcription. Mechanisms of tight control of TERT by TL shown here are likely to have major implications in telomere-related physiologies, particularly, cancer, ageing, and pluripotency.
- 30 September 2025
- Published by eLife Sciences Publications, Ltd in eLife
- Vol. 14
Open AccessTelomeres control human telomerase (TERT) expression through non-telomeric TRF2
The function of the human telomerase reverse transcriptase (referred hereafter as TERT) in the synthesis and maintenance of chromosome ends, or telomeres, is widely understood. Whether and how telomeres, on the other hand, influence TERT regulation is relatively less studied. We found TERT was transcriptionally altered depending on telomere length (TL). This resulted from TL-dependent binding of TRF2 between telomeres and the TERT promoter. TERT promoter-bound TRF2 was non-telomeric and did not involve the looping of telomeres to the TERT promoter. Cell lines from different tissue types fibrosarcoma (HT1080), colon cancer (HCT116), and breast cancer (MDA-MB-231), engineered for either telomere elongation/shortening, gave an increase/decrease in TERT, respectively. Mechanistically, we show TERT promoter-bound non-telomeric TRF2 recruits the canonical PRC2-complex, inducing repressor histone H3K27-trimethylation in a TL-dependent fashion. This was further supported by TL-dependent promoter activity from an exogenously inserted TERT reporter. Increase in TL over days followed by a gradual decline, resulted in activation followed by repression of TERT in a concerted manner, further implicating TL as a key factor for TERT regulation. Notably, on reprogramming primary fibroblasts to induced pluripotent stem cells (iPSCs), TRF2 loss from the TERT promoter was evident along with telomere elongation and TERT upregulation. Conversely, on telomere shortening in iPSCs, TERT promoter-bound TRF2 was restored with a marked reduction in TERT, further supporting the causal role of TL in TERT transcription. Mechanisms of tight control of TERT by TL shown here are likely to have major implications in telomere-related physiologies, particularly, cancer, ageing, and pluripotency.