iiinova — where your digital self keeps evolving

Floris Barthel @florisbarthel.bsky.social · 06/10/2026

In mice with orthotopic GBM43 tumors, plasma cfDNA was elevated over PBS-injected controls. Separating human from mouse reads, we found tumor-derived fragments were shorter than host fragments, and their copy number profile recapitulated the parental chr7 and chr9 gains.

Genome-wide copy number heatmap with chromosomes 1 to 22, X and Y along the top. The first row, GBM43 genomic DNA, shows gains of chromosomes 7, 8 and 9. Plasma cfDNA from GBM43 tumor-bearing mice, treated with vehicle or temozolomide, shows matching gains of chromosomes 7 and 9, outlined in red. Plasma from PBS-injected control mice shows no such gains.

Box plots of mean cfDNA fragment size (0 to 300 bp) for mouse-derived (red) and human-derived (blue) reads. In tumor-bearing mice, human fragments are shorter than mouse fragments (P = 0.039). In PBS-injected control mice, the difference is not significant (P = 0.547).

100

Floris Barthel @florisbarthel.bsky.social · 06/10/2026

Temozolomide (TMZ) changed the picture. In treated GBM43, cfDNA yield rose sharply and correlated with dead cell counts (R = 0.85), no longer with live cells (R = −0.04). Treated cfDNA showed nucleosomal laddering and a narrower pool of detectable variants.

Two scatter plots for temozolomide-treated GBM43 cultures, with points colored by replicate. cfDNA yield correlates strongly with dead cell counts (R = 0.85, P < 0.001) and does not correlate with live cell counts (R = −0.04, P = 0.896).

Line graph of cfDNA yield from 24 to 120 hours. Temozolomide-treated GBM43 cultures rise sharply after 48 hours to about 350 ng at 120 hours. Untreated cultures increase gradually to about 120 ng.

100

Floris Barthel @florisbarthel.bsky.social · 06/10/2026

Fragment sizes told the same story. Astrocyte cfDNA showed a mononucleosomal peak (~185 bp) typical of apoptosis, GBM43 cfDNA lacked nucleosomal peaks, and co-cultures produced a multimodal pattern (174, 330, 486 bp) seen in neither monoculture.

Three TapeStation electropherograms of cfDNA fragment size at 120 hours. GBM43 monoculture shows no nucleosomal peaks, with signal concentrated in long fragments. GBM43 plus astrocyte co-culture shows a multimodal pattern with peaks at 174, 330 and 486 bp. Astrocyte (NHA) monoculture shows a dominant mononucleosomal peak at 185 bp.

100

Floris Barthel @florisbarthel.bsky.social · 06/10/2026

Over 5 days, GBM43 outgrew the astrocytes and astrocyte death increased. In co-culture cfDNA, NHA-specific VAFs rose 2.4% per day while GBM43-specific VAFs fell 3.6% per day. Dying astrocytes diluted the tumor signal, much as host DNA does in patient plasma.

Three line graphs of mean VAF from 24 to 120 hours for shared, GBM43-specific and NHA-specific variants. In GBM43 monoculture cfDNA, GBM43-specific VAF rises (P = 0.02). In NHA monoculture cfDNA, VAFs stay flat. In co-culture cfDNA, NHA-specific VAF rises from about 35% to about 47% (P = 0.0008) while GBM43-specific VAF falls from about 23% to about 7% (P < 0.001). Shared variants stay near 80% throughout.

100

Floris Barthel @florisbarthel.bsky.social · 06/10/2026

To model the microenvironment, we co-cultured GBM43 with normal human astrocytes (NHAs). Variants unique to each cell type let us attribute cfDNA to its source: GBM43 variants were absent from astrocyte cfDNA, and NHA variants from tumor cfDNA.

Two heatmaps of variant allele frequency in cfDNA from GBM43 plus astrocyte co-culture, GBM43 monoculture and astrocyte (NHA) monoculture, each at 24 to 120 hours. Top: variants unique to GBM43 appear in GBM43 and co-culture cfDNA and are absent from NHA cfDNA. Bottom: variants unique to NHA appear in NHA and co-culture cfDNA and are absent from GBM43 cfDNA. Cells show supporting read counts.

100

Floris Barthel @florisbarthel.bsky.social · 06/10/2026

We expected cfDNA to come mostly from dying cells. Instead, cfDNA yield correlated more strongly with live cell counts than with dead cell counts (GBM43: R = 0.88 vs 0.47). So cfDNA release in these models is not driven by apoptosis alone.

Four scatter plots of cfDNA yield against cell counts. GBM43, left: yield correlates weakly with dead cell counts (R = 0.47, P = 0.088) and strongly with live cell counts (R = 0.88, P < 0.001), with points colored by replicate. GBM12, right: yield correlates with dead cell counts (R = 0.85, P = 0.071, not significant) and with live cell counts (R = 0.92, P = 0.027).

100

Floris Barthel @florisbarthel.bsky.social · 06/10/2026

We cultured two patient-derived GBM lines, GBM43 and GBM12, and sampled the media every 24 h for 5 days to measure cell counts, cfDNA yield, fragment size and variants. All variants in each line's genomic DNA were detected in its cfDNA.

Two panels of variant allele frequency (VAF) in cfDNA over five days. Left, GBM43: mean VAF rises from about 55% at 24 hours to about 80% at 120 hours (P = 0.026), above a heatmap of six GBM43 variants including TP53, SUFU and NF1. Right, GBM12: mean VAF stays near 55% at all timepoints (P = 0.81), above a heatmap of seven GBM12 variants including two in EGFR. Heatmap cells show supporting read counts.

100

Floris Barthel @florisbarthel.bsky.social · 06/10/2026

Glioblastoma is monitored mostly by MRI, which can miss early changes. Cell-free DNA (cfDNA) in blood is a promising alternative, but we know little about what governs its release. Our new paper in Neuro-Oncology, led by @sharvarimankame.bsky.social, dissects it. 🧵 doi.org/10.1093/neuonc/noag204

Top: a five-step workflow schematic: cfDNA extraction from GBM43 and GBM12 cultures sampled at 24, 48, 72, 96 and 120 hours, library preparation, targeted sequencing, quality control, and variant calling. Bottom: brightfield micrographs of the two patient-derived glioblastoma lines. GBM43 grows as adherent, spindle-shaped cells and GBM12 as rounded neurospheres (scale bars 300 µm).

150

Floris Barthel @florisbarthel.bsky.social · 28/07/2026

Feature sets are not tied to a genome: any annotation that tiles a reference can become one. Here is a database we built for the Arabidopsis Col-CEN T2T reference, with chromosome, gene, region, and repeat feature sets, including the CEN180 satellite arrays.

KaryoScope annotation for the arabidopsis genome shown across four feature sets

111

Floris Barthel @florisbarthel.bsky.social · 28/07/2026

Some run statistics: peak 941% CPU across 10 threads, peak 10.2 GB RAM on an M1 Max. Broad plateaus are k-mer lookup, sharp bursts are hierarchy-aware smoothing, and each of the six feature sets appears as its own block.

CPU and RAM statistics for KaryoScope annotating the HG002 genome on a MacBook M1 laptop

110

Floris Barthel @florisbarthel.bsky.social · 28/07/2026

Excited to share a big KaryoScope update this past week: the HKS k-mer backend is now integrated, and you can build databases for any features of interest. The complete @humanpangenome.bsky.social HG002 assembly (~6.3 Gbp) now annotates in 21 minutes on a MacBook Pro. 🧵

Macbook laptop showing KaryoScope actively running alongside the final output karyotype plots

193

Floris Barthel @florisbarthel.bsky.social · 22/05/2026

We expected broad genomic instability. Instead, telomere crisis zeroed in on the acrocentrics, consistent with findings from Beth Sullivan's lab over a decade ago using dominant-negative TRF2. doi.org/10.1371/jour... doi.org/10.1371/jour...

110

Floris Barthel @florisbarthel.bsky.social · 22/05/2026

To probe nuclear organization, we used Hi-C. Standard pipelines discard the rare inter-chromosomal contacts at acrocentric arms, so we recovered them with KaryoScope, our alignment-free tool. These contacts were persistently depleted in crisis cells. bsky.app/profile/flor...

KaryoScope Hi-C inter-chromosomal contact maps and the log2 fold-change map with the chr13 row/column in blue

121

Floris Barthel @florisbarthel.bsky.social · 22/05/2026

Why acrocentrics? Their short arms carry the NORs that nucleate the nucleolus. In crisis cells, nucleoli broke down: compact spheres gave way to dispersed, necklace-like structures, with rDNA transcription and processing decoupled.

UBF/fibrillarin immunofluorescence, compact vs necklace nucleoli

110

Floris Barthel @florisbarthel.bsky.social · 22/05/2026

The rearrangements were strikingly dynamic. Chromosome 13 translocations were common early but transient, replaced by new partners at later passages, while whole-chromosome losses persisted. Crisis generates a shifting, subclonal landscape of damage.

DAPI metaphase spreads with arrowheaded translocations, plus pie charts of shifting translocation composition

110

Floris Barthel @florisbarthel.bsky.social · 22/05/2026

Multicolor FISH revealed subclonal abnormalities largely invisible to short-read sequencing. The bias was striking: acrocentric chromosomes carried 41% of all abnormalities despite being 15% of the autosomal genome. Chr13 was abnormal in >92% of metaphases.

chromosomal abnormality catalog organized by type (deletions / translocations / fusions) across four clones showing an abundance of acrocentric involvement

110

Floris Barthel @florisbarthel.bsky.social · 22/05/2026

To capture it, we grew normal human astrocytes with HPV E6/E7 (inactivating p53 and Rb) but no telomerase, so cells divide through progressive telomere erosion. We tracked them longitudinally from senescence bypass through crisis, an otherwise inaccessible window.

the longitudinal growth curve showing the senescence and crisis growth plateaus

110

Floris Barthel @florisbarthel.bsky.social · 22/05/2026

Telomere crisis is an engine of genomic instability, driving the structural evolution of cancer genomes. Our new preprint finds this damage isn't random: it converges on the nucleolus and the chromosomes that build it. 🧵 www.biorxiv.org/content/10.6...

two multicolor FISH karyograms of crisis cells

151

Floris Barthel @florisbarthel.bsky.social · 17/05/2026

KaryoScope analyzes sequence data faster than instruments from @pacbio.bsky.social , @nanoporetech.com, Illumina, and @elembio.bsky.social (and others) produce it, on laptop-grade hardware. Real-time, on-instrument sequence annotation is within reach.

Four-panel performance scaling of KaryoScope across CPU thread counts. Top left: feature-ID lookup throughput (Mb/s) for KMC (red) and HKS (blue) k-mer backends. Bottom left: peak memory (GB) for each backend. Top right: smoothing-step runtime (minutes) comparing Python (red) and Rust (blue) implementations. Bottom right: smoothing-step memory. HKS and Rust each scale substantially better than KMC and Python.

220

Floris Barthel @florisbarthel.bsky.social · 17/05/2026

KaryoScope works on any sequence input, beyond diploid assemblies: long reads, short reads, Hi-C, RNA-seq, metagenomics. It detects SVs from individual long reads (manuscripts forthcoming) and dissects cancer genome assemblies (HG008): bsky.app/profile/florisbarthel.bsky.social/post/3mle3v75bpk2k

KaryoScope karyotype of the HG008 pancreatic tumor assembly. Each chromosome is shown as a track colored by chromosome of origin, revealing complex rearrangements: derivative chromosomes from reciprocal translocations, dicentric chromosomes, and chromoplexy-like patterns where blocks of one chromosome's sequence appear within another.

100

Floris Barthel @florisbarthel.bsky.social · 17/05/2026

@khmiga.bsky.social and @glennislogsdon.bsky.social have charted centromeres. KaryoScope adds pangenome-scale variation: chr9 megabase inversion, chr3/chr5 repeat losses, FISH-validated. See also: www.science.org/doi/10.1126/... www.biorxiv.org/content/10.6...

120

Floris Barthel @florisbarthel.bsky.social · 17/05/2026

FSHD1, a muscular dystrophy, is caused by structural changes in D4Z4, a complex subtelomeric repeat array on chromosomes 4q and 10q. Across hundreds of @humanpangenome.bsky.social, KaryoScope first catalogs D4Z4 diversity, including configurations previously not described.

Pangenome-scale D4Z4 macrosatellite catalog from KaryoScope. Panels A and C–F: KaryoScope tracks (satellites, genes, repeats, subtelomere, D4Z4 arrays) for representative haplotypes across chromosomes 4q and 10q, including canonical, multi-array, and mixed configurations. Waffle charts (B) show D4Z4 structural category frequencies across 249 chr4 and 333 chr10 haplotypes. Contingency tables (G) summarize diploid combinations. Scatterplot (H) compares repeat-unit counts between haplotypes within individuals.

110

Floris Barthel @florisbarthel.bsky.social · 17/05/2026

Robertsonian translocations are complex chromosomal rearrangements of acrocentric chromosomes. @aphillippy.bsky.social, @thinks.lol, and Jen Gerton showed SST1 is the fusion substrate; KaryoScope confirms this from k-mers alone. See also: www.nature.com/articles/s41... www.nature.com/articles/s41...

KaryoScope annotation of three Robertsonian translocation assemblies (top: rob(14;21), middle: rob(13;14) in GM03786, bottom: rob(13;14) in GM04890). Each row shows six progressively zooming panels from full contig (~130–200 Mb) down to 200 kb at the fusion site. Tracks per panel show chromosome of origin, genes, repeats, satellites, and acrocentric features. The SST1 satellite block (green) is visible at every fusion point.

130

Floris Barthel @florisbarthel.bsky.social · 17/05/2026

Built on k-mer matching (short, fixed-length DNA fragments), KaryoScope annotates a complete diploid human genome at base-pair resolution in ~2 minutes, across repeats, satellite families, genes, and chromosome-end structure. ~300× faster than RepeatMasker.

Schematic comparing FISH (fluorescence in situ hybridization) and KaryoScope workflows side by side in four parallel steps: feature definition (cloning vectors vs. genomic feature sets), fragmentation (vector processing vs. k-mer enumeration), label assignment (fluorescent probes vs. labeled k-mer sets), and query (in situ probe binding vs. k-mer database lookup). KaryoScope is presented as a computational analog of FISH.

120

Floris Barthel @florisbarthel.bsky.social · 17/05/2026

The human genome's most variable and clinically important regions (centromeres, telomeres, and acrocentric short arms) have been hardest to study at scale. Thrilled to share KaryoScope, our new preprint that brings them within reach. 🧵 www.biorxiv.org/content/10.6...

KaryoScope karyotype of the HG002 diploid assembly. All 22 autosomes plus X and Y, each shown as paired haplotypes (h1, h2). Each chromosome has a full-length track colored by chromosome of origin, a centromere zoom panel showing satellite composition, and a subtelomere zoom panel. Legends map chromosomes, satellite families, and subtelomeric features to colors.

35623

Floris Barthel @florisbarthel.bsky.social · 08/05/2026

KaryoScope also resolved two reciprocal translocations (chr12↔15, chr18↔X), foldback inversions, acrocentric short-arm rearrangements, and complex multi-chromosomal events; all reconciled against dGH ground truth. Check out the pre-print for a complete overview of everything we observed.

KaryoScope visualization of two reciprocal translocations

100

Floris Barthel @florisbarthel.bsky.social · 08/05/2026

That α-satellite fusion produced what turned out to be a putative functional dicentric chromosome: two CDRs ~1 Mbp apart on D7Z1 and D6Z1, beautifully characterized by CenMAP from the @glennislogsdon.bsky.social lab. 🔗 github.com/logsdon-lab/...

CenMAP visualization of the 7p-6p dicentric chromosome

100

Floris Barthel @florisbarthel.bsky.social · 08/05/2026

KaryoScope resolved a striking chromoplexy event linking chr 3, 6, 7, and 11 into three hybrid tumor chromosomes, and pinpointed the chr6↔chr7 breakpoint deep inside the α-satellite array of both centromeres.

Chromoplexy causing a complex series of events linking chromosomes 3, 6, 7, and 11 in three hybrid tumor chromosomes

100

Floris Barthel @florisbarthel.bsky.social · 08/05/2026

Our lab contributed via KaryoScope, our k-mer-based sequence annotation tool. It produced an ISCN-annotated karyotype of the tumor assembly that was concordant with KromaTiD dGH SCREEN cytogenetics across all 35 tumor chromosomes.

Side-by-side computational and traditional multi-color karyotype for HG008T

100

Floris Barthel @florisbarthel.bsky.social · 08/05/2026

🧵 Excited to contribute to a new preprint led by Justin Zook & the NIST/GIAB team: "A complete human pancreatic cancer genome": the first near-complete, haplotype-resolved tumor cell line assembly (HG008-T, hypodiploid PDAC). 📄 www.biorxiv.org/content/10.6...

2102

Floris Barthel @florisbarthel.bsky.social · 23/04/2026

Somatic SV breakpoints were enriched at telomeres and centromeres regardless of local telomere length, a constitutive layer of structural fragility replicated in external genomic datasets. This work was led by Maryam in the lab with a great team of collaborators at Mayo Clinic, NCI and Weizmann.

000

Floris Barthel @florisbarthel.bsky.social · 23/04/2026

Our working model: short telomeres drive breakage via BFB cycles; long ALT-maintained telomeres coincide with ecDNA, potentially linked to replicative repair mechanisms.

110

Floris Barthel @florisbarthel.bsky.social · 23/04/2026

New preprint! We used Oxford Nanopore long-read sequencing to ask: does chromosome-arm-specific telomere length shape how IDH-mutant astrocytoma genomes evolve? Short answer: yes, and in two opposite directions. 🔗 www.biorxiv.org/content/10.6...

122

Floris Barthel @florisbarthel.bsky.social · 17/10/2025

Yi-An Chen presenting his work on the telomeric interactome #ASHG2025, now also on biorxiv: www.biorxiv.org/content/10.1...

030

Floris Barthel @florisbarthel.bsky.social · 16/10/2025

Ogechukwu (Mimi) Mbegbu, super star RA and resident FISH-ing expert showcasing her cytogenetic analysis of astrocytes undergoing telomere dysfunction at #ASHG2025. Lots of chromosomal chaos observed and a 3-fold bias towards acrocentric abnormalities?! Stay tuned for the mechanism

020

Floris Barthel @florisbarthel.bsky.social · 14/10/2025

We validated these contacts with classic 3C/PCR and confirmed the 3D proximity using Southern Blotting and stunning 3D FISH data, proving these are real, higher-order chromatin structures.

100

Floris Barthel @florisbarthel.bsky.social · 14/10/2025

Specifically, we found strong enrichment at three classes of repetitive elements: Interstitial Telomeric Sequences (ITS), Telomere-Associated Repeat 1 (TAR1), and D20S16 elements.

110

Floris Barthel @florisbarthel.bsky.social · 14/10/2025

The telomeric interactome spans the entire genome! While contacts are enriched near telomeres (<100 Kb), over 80% of all interactions occurred >5 Mb away from chromosome ends. This isn't just about the subtelomere.

100

Floris Barthel @florisbarthel.bsky.social · 14/10/2025

Our initial goal was to map telomeric regulation of key genes, like TERT. The reality? Gene-level interactions were massively outnumbered by contacts with repetitive elements! 🤯

100

Floris Barthel @florisbarthel.bsky.social · 14/10/2025

To map the telomeric interactome, we developed Telomere-C (Telomere Conformation Capture). This novel, ligation-free method uses a biotinylated PNA probe to capture telomeric DNA, allowing us to measure nuclear contacts efficiently using short-read sequencing.

100