ASH 2025: AI uncovers how DNA architecture failures trigger blood cancer
Presented at the American Society of Hematology annual meeting, new data from the University of Miami Miller School of Medicine put genome architecture squarely in the cancer conversation. The takeaway is blunt: the way DNA folds can silence tumor suppressor programs even when the DNA sequence is intact.
The team introduced a practical concept for hematology and genomics labs: architectural tumor suppression. Cohesin and CTCF do more than organize chromatin-they keep enhancer-promoter loops intact so tumor suppressors stay on.
The study at a glance
Title: "SMC3 and CTCF Haploinsufficiency Drive Lymphoid Malignancy via 3D Genome Dysregulation and Disruption of Tumor Suppressor Enhancer-Promoter Loops."
Institution: Sylvester Comprehensive Cancer Center, University of Miami Miller School of Medicine.
Core concept: architectural tumor suppression
Partial loss of SMC3 or CTCF doesn't melt the genome. It selectively erodes short-range enhancer-promoter contacts-the precise wiring that keeps tumor suppressor genes active.
When those short loops fray, transcriptional support for genes such as Tet2, Kmt2d, and Dusp4 drops. In B-cells, that bottlenecks fate decisions and blocks maturation into plasma cells, increasing lymphoma risk.
Methods that made this possible
- Hi-C maps to resolve 3D chromatin contacts, with emphasis on short-range loops.
- Single-cell RNA-seq to track lineage choices and state transitions.
- Epigenetic profiling to mark enhancer activity and accessibility.
- AI-driven integration to expose cross-modal patterns invisible in single datasets.
"We've long known that mutations drive cancer," said Dr. Martin Rivas of Sylvester Comprehensive Cancer Center. "But this work shows that architecture-the way DNA folds-can be just as important. It's like losing the blueprint for a building while construction is under way."
Key findings for researchers
- Selective loop loss: Haploinsufficiency of SMC3 or CTCF disproportionately weakens short enhancer-promoter loops that sustain tumor suppressor expression.
- Cell-fate impact: B-cells stall before plasma cell maturation, creating a permissive state for malignancy.
- Architecture → expression link: AI analysis mapped loop erosion to reduced transcription at tumor suppressor loci, clarifying causal direction.
- Clinical signal: In DLBCL cohorts, lower SMC3 expression associates with poorer outcomes, supporting genome architecture as a potential prognostic biomarker.
"AI allowed us to see patterns invisible to the human eye-how losing just one copy of a gene reshapes the entire 3D landscape," Rivas noted.
Why this matters for your program
- Study design: If you model lymphoma or B-cell biology, measure short-range looping directly (Hi-C/HiChIP/PLAC-seq) around tumor suppressor loci, not just global compartments or TADs.
- Haploinsufficiency is enough: Track one-allele losses and expression reductions in cohesin/CTCF pathway genes; subtle dosage shifts have system-level effects.
- Analysis stack: Integrate chromatin contact maps with single-cell state trajectories to quantify how architecture gates lineage decisions.
- Translational angle: Evaluate SMC3/CTCF levels and loop integrity as biomarkers in DLBCL; explore loop-restoration strategies or enhancer reactivation as therapeutic hypotheses.
Practical next steps
- Prioritize enhancer-promoter loop QC in pipelines; flag short-loop attrition near tumor suppressor hubs.
- Co-sample scRNA-seq and chromatin contact data in the same models or matched patients where feasible.
- Benchmark AI integrators on multi-omic ground truth; stress-test against batch, cell-cycle, and coverage artifacts.
- Correlate architectural metrics with clinical endpoints to refine risk stratification in DLBCL.
Quotes worth bookmarking
"This is where computational biology shines," said Rivas. "AI allowed us to see patterns invisible to the human eye-how losing just one copy of a gene reshapes the entire 3D landscape."
"We're entering an era where cancer treatment could mean repairing architecture, not just fixing broken genes," Rivas added.
Session details
Scientific Program: Enhancer Function in Lymphoid Development and Lymphoma
Presentation: Dec. 6, 2025, 9:50-10:10 a.m. EST, OCCC - W312
Further reading
AI literacy for multi-omics teams
If you're building internal capacity for AI-first analysis of Hi-C, scRNA-seq, and epigenomics, consider structured upskilling resources.
Your membership also unlocks:
