Systematic multivariate analysis of chromatin complex dependencies reveals Set1C/COMPASS as a melanoma-enriched epigenetic vulnerability
Abstract
Epigenetic dysregulation is a common feature of cancer. It creates vulnerabilities arising from an increased reliance on chromatin-based mechanisms that sustain malignant transcriptional states. While many chromatin regulators are broadly required for cellular viability, others function in a context-dependent manner across distinct oncogenic settings, tissue lineages, and differentiation states. Moreover, chromatin regulators often operate within multi-subunit complexes; thus, epigenetic vulnerabilities emerge from coordinated complex activities. Here, we integrate large-scale genetic dependency maps from human cancer cell lines with curated epigenetic complex annotations to perform a systematic, multivariate analysis of complex-level epigenetic dependencies across cancer lineages. Our analysis reveals that dependencies frequently cluster among functionally related chromatin complexes and that biologically related cancer types (e.g., hematologic malignancies) share similar dependency patterns, consistent with shared underlying epigenetic requirements. Focusing on melanoma, we identify multiple enriched epigenetic complex dependencies, including complexes previously associated with recurrent genetic alterations or melanocyte lineage regulation, as well as a previously unrecognized vulnerability involving the H3K4 methyltransferase complex Set1C/COMPASS. This dependency is not restricted to a specific melanoma differentiation state, but genetic depletion of CXXC1 (a critical complex-specific subunit) suggests that CXXC1-dependent melanoma cells require Set1C/COMPASS activity to maintain global H3K4 trimethylation (H3K4me3) and proliferation. Integrative modeling links Set1C/COMPASS dependency to MYC- and E2F-driven transcriptional programs, which are suppressed upon complex inhibition. Together, this work combines integrative, multivariate analysis of lineage-enriched epigenetic dependencies with genetic perturbation, transcriptional profiling, and single-cell analysis to uncover an enriched epigenetic dependency on Set1C/COMPASS in melanoma cells. Author SummaryCancer cells often rely on abnormal regulation of gene activity to support uncontrolled growth and survival. This regulation is controlled not only by genetic mutations, but also by epigenetic mechanisms, chemical and structural modifications to DNA and its associated proteins that determine which genes are turned on or off. Several therapies that target epigenetic regulators have shown promise, particularly in blood cancers. However, identifying which epigenetic mechanisms are most important in specific cancers remains challenging, especially because epigenetic regulators frequently work together as multi-protein complexes. In this study, we combine large-scale public datasets across many cancer lineages with computational modeling to systematically identify lineage-enriched epigenetic vulnerabilities. We found that certain epigenetic complexes are selectively important in specific cancer lineages. In melanoma, an aggressive skin cancer, we identified a previously unrecognized dependence on a protein complex that modifies chromatin at gene promoters. We show that disrupting this complex can impair gene programs that drive cell division and block cancer cell growth. Moreover, our study demonstrate how integrative computational approaches combined with experiments can uncover new targets for potential cancer therapy studies.
02 /Specification & Classification
Click any tag to browse matching papers- Citations
- 0
- FWCI
- 0.00
- Field percentile
- 8th percentile
- Organism
- Assay
- Research Focus
- Approach
In-depth summary
Claude Opus 5.5 · xhigh effort · 8 Nov 2026 · from the full text of version 2 (21 Jul 2026) · CC BY 4.0
Complex-level DepMap dependency modeling across 42 cancer types nominates Set1C/COMPASS as melanoma-enriched; CXXC1 knockdown lowers H3K4me3, curbs proliferation, and suppresses MYC/E2F programs in responsive melanoma lines.
Earlier melanoma epigenetic work examined individual regulators or single classes of chromatin modifiers, although chromatin regulators act as multi-subunit complexes and no lineage- or differentiation-state-resolved map of complex-level dependencies had been built. Combining DepMap dependencies, CCLE transcriptomes and curated complex annotations, this work nominates and experimentally probes a previously unreported Set1C/COMPASS dependency in melanoma.
Key results
ShownAveraging DepMap CRISPR dependencies over curated chromatin complexes across 42 cancer lineages nominated several melanoma-enriched complexes, including Set1C/COMPASS; gene-level testing independently flagged its subunits CXXC1 and SETD1B.
Complex-level limma one-versus-rest (43 complexes, 216 genes, exploratory FDR ≤ 0.1) flagged Set1C/COMPASS, G9a/GLP, NURF, HBO1, ATRX-DAXX, Elongator and BRCA1-A in melanoma; at gene level, 26 of 735 regulators had melanoma-elevated dependency (difference > 0.1, FDR ≤ 0.05), including CXXC1 and SETD1B, both higher in 67 melanoma vs 1,040 other lines. · Epigenetic dependency analysis of cancer cell…; Pan-cancer complex dependency analysis from…; Figure legends
ShownCXXC1 siRNA depleted CXXC1 protein in all five tested BRAF-mutant lines but lowered global H3K4me3 and growth only in a subset; in MALME3M and UACC62, H3K4me3 loss co-occurred in single cells with Ki-67 loss, and MALME3M cells accumulated in G0/G1.
Knockdown ~75% to 95%; H3K4me3^High fraction fell >30% (FDR < 0.05) only in MALME3M, UACC62, LOXIMVI; cell count fell ~21% and ~53% in MALME3M and UACC62 (FDR < 0.05), not in SKMEL28 or WM115; interaction P ≈ 0.03 (H3K4me3), P ≈ 0.002 (count); Ki-67^High fell ~50% in MALME3M and UACC62, where CXXC1/H3K4me3 double-high cells (~80-90%) became ~60-80% double-low. · CXXC1 depletion reduces H3K4me3 in…; Single-cell analysis confirms coupling of…; Statistical analysis and reproducibility
ShownBulk RNA-seq after CXXC1 knockdown revealed a shared response in the three H3K4me3-responsive lines, absent from SKMEL28, dominated by suppression of MYC-, E2F- and G2-M-linked cell-cycle programs and induction of p53 pathway genes.
Four lines in biological triplicate (12 samples); PC4 separated si-CXXC1 from si-NTC in MALME3M, UACC62 and LOXIMVI but not SKMEL28; consensus of 231 down- and 241 up-regulated genes; Enrichr top downregulated terms G2-M checkpoint and MYC targets, top upregulated term p53 pathway; fgsea on mean DESeq2 Wald statistics again highlighted E2F, G2-M and MYC sets. · CXXC1 depletion suppresses MYC- and E2F-linked…; Analysis of gene expression data following CXXC1…; Figure legends
ShownSet1C/COMPASS dependency was not confined to melanocytic-like or undifferentiated/neural crest-like melanoma lines, even though epigenetic dependency profiles overall did discriminate the two states.
Gaussian mixture classification of 71 CCLE lines by Tsoi signatures; PLS-DA on 735 regulators in 51 lines (34 melanocytic-like vs 17 undifferentiated) gave cross-validated ROC AUC 0.82 ± 0.14 and 78 melanocytic- vs 178 undifferentiated-associated genes (|VIP| ≥ 1); among Set1C/COMPASS subunits only ASH2L dependency differed by state (Wilcoxon). · Melanoma-enriched Set1C/COMPASS dependency is…; Differentiation state classification of melanoma…; Partial least squares discriminant analysis…
SuggestedAcross melanoma lines, higher Set1C/COMPASS dependency tracked with expression of and co-dependency on MYC and E2F target programs, and this co-dependency pattern recurred in most other cancer lineages.
Pre-ranked fgsea over 51 CCLE melanoma lines vs mean dependency of CXXC1, SETD1B, ASH2L, DPY30, RBBP5, WDR5, WDR82: positive at both levels for MYC and E2F targets, DNA repair and oxidative phosphorylation; epithelial-mesenchymal transition, KRAS and IL-2/STAT5 negative; lineage-wise analysis found MYC/E2F enrichment in most lineages incl. B-ALL, HNSC, AML. · Set1C/COMPASS dependency associates with MYC-…; CXXC1 depletion suppresses MYC- and E2F-linked…; Figure legends
Methods, data and reproducibility
Mean DepMap 25Q2 CRISPR dependency probabilities over curated EpiFactors complex subunits were tested per lineage with limma one-versus-rest, and melanoma differentiation states compared via PLS-DA; CXXC1 siRNA in melanoma lines was read out by 4i multiplexed imaging and bulk RNA-seq (DESeq2, fgsea).
- DepMap Public 25Q2 CRISPR gene dependency probabilities (Chronos-based) for CCLE cancer cell lines, lineages with at least five lines retained · human · n = >1,000 cell lines; 42 of 76 lineages retained
- Melanoma vs non-melanoma DepMap comparison of CXXC1 and SETD1B dependency · human · n = 67 melanoma vs 1,040 non-melanoma cell lines
- CCLE melanoma RNA-seq scored with Tsoi et al. differentiation-state gene signatures · human · n = 71 cell lines
- CCLE melanoma lines with matched RNA-seq and DepMap dependencies (PLS-DA and Set1C/COMPASS GSEA) · human · n = 51 cell lines (34 melanocytic-like; 17 undifferentiated)
- EpiFactors epigenetic regulators and manually collapsed chromatin complex reference set · n = 735 genes; 43 complexes spanning 216 genes
- BRAF-mutant melanoma lines (LOXIMVI, MALME3M, SKMEL28, UACC62, WM115) treated with CXXC1 or control siRNA for 96 h, profiled by 4i imaging · human · n = five cell lines; n = 3 replicates per condition
- Bulk RNA-seq after CXXC1 or non-targeting siRNA in MALME3M, UACC62, LOXIMVI and SKMEL28 · human · n = 12 samples (four lines, biological triplicate) · GSE324132
- Methods & tools
- DepMap / CCLE (Chronos CRISPR dependency scores), EpiFactors, limma (one-versus-rest, empirical Bayes), PLS-DA with VIP scores, mclust (Gaussian mixture models), Enrichr, fgsea with MSigDB Hallmark, GO:BP, KEGG, DESeq2, STAR and RSEM (GRCh38, GENCODE v41), FastQC, Trim Galore, Qualimap, 4i iterative indirect immunofluorescence (Operetta CLS) with ON-TARGETplus siRNA SMARTpool, CellProfiler v4.2.8 and ImageJ v2.17.0
- Data & accessions
- https://www.ncbi.nlm.nih.gov/geo/https://github.com/fallahi-sichani-lab/chromatin-complex-dependencies-Set1C-COMPASS-melanoma
- Software & environment
- R (version 4.x); ImageJ v2.17.0; CellProfiler v4.2.8; STAR and RSEM against GRCh38 with GENCODE v41; DepMap Public 25Q2 (CRISPRGeneDependency.csv)
- Reproducibility gaps
- Further information needed to reanalyze the data is available only on request; R version given only as 4.x, and versions of limma, DESeq2, fgsea and mclust are not stated.
Limitations (6)
- No rescue experiments with wild-type CXXC1 or other Set1C/COMPASS subunits such as SETD1B were done, so off-target effects of the CXXC1 siRNA pool on the growth defect cannot be excluded. · stated by the authors
- The data cannot tell whether Set1C/COMPASS controls MYC and E2F targets directly through promoter H3K4me3 or indirectly; chromatin occupancy profiling (ChIP-seq or CUT&RUN for CXXC1 and H3K4me3) is still needed. · stated by the authors
- What predicts Set1C/COMPASS dependency, and why its prevalence differs across lineages, remains unknown; subunit mutations or expression levels did not explain it. · stated by the authors
- In vivo relevance and the therapeutic value of pharmacologic Set1C/COMPASS inhibition were not tested. · stated by the authors
- The complex-level screen used a permissive exploratory FDR ≤ 0.10, because averaging over all subunits can dilute effects; its hits needed gene-level corroboration. · stated by the authors
- Perturbation work covers only five BRAF-mutant lines with three wells per arm and a single nonresponsive RNA-seq control (SKMEL28); LOXIMVI lost H3K4me3 without a significant growth drop (FDR = 0.09), so the H3K4me3-growth link rests on few lines. · noted in this summary
Who should read:Read this if you study melanoma epigenetics or want a template for mining DepMap dependencies at the level of chromatin complexes rather than single genes, with imaging and RNA-seq follow-up.
Reproducibility ledger
Nobody has reported trying to reproduce this paper yet. Did you replicate an experiment or run the code? Your report saves the next lab a week.
Sign in to file a reproducibility report.
Related papers
- Functional mapping of epigenetic regulators uncovers coordinated tumor suppression by the HBO1 and MLL1 complexes
Yuning J. Tang, Haiqing Xu, … Monte M. Winslow·Cancer DiscoveryPeer-reviewed·Sep 25, 2025
- Chromatin Landscape of Cancer Cell Lines Identifies Enhancer Subtypes
Mayinuer Maitituoheti, Emre Arslan, … Kunal Rai·bioRxiv·Aug 4, 2026
- A network of epigenomic and transcriptional cooperation encompassing an epigenomic master regulator in cancer
Stephen P. Wilson, Fabian V. Filipp·npj Systems Biology and ApplicationsPeer-reviewed·Jun 7, 2018
- Menin-MLL1 complex cooperates with NF-Y to promote HCC survival
Margarita M. Dzama, Mallory Sokolowski, … Jesse R. Raab·Cell ReportsPeer-reviewed·Nov 25, 2025
- Epigenetic dysregulation from chromosomal transit in micronuclei
A. Agustinus, R. Raviram, … S. F. Bakhoum·bioRxiv·Jan 12, 2022
- Bivalent and Broad Chromatin Domains Regulate Pro-metastatic Drivers in Melanoma
C. Terranova, M. Tang, … K. Rai·bioRxiv·Aug 1, 2019