Tumor evolution is driven by various mutational processes, ranging from single-nucleotide variants (SNVs) to large structural variants (SVs) to dynamic shifts in DNA methylation. Current short-read sequencing methods struggle to accurately capture the full spectrum of these genomic and epigenomic alterations due to inherent technical limitations. To overcome that, here we introduce an approach for long-read sequencing of single-cell derived subclones, and use it to profile 23 subclones of a mouse melanoma cell line, characterized with distinct growth phenotypes and treatment responses. We develop a computational framework for harmonization and joint analysis of different variant types in the evolutionary context. Uniquely, our framework enables detection of recurrent amplifications of putative driver genes, generated by independent SVs across different lineages, suggesting parallel evolution. In addition, our approach revealed gradual and lineage-specific methylation changes associated with aggressive clonal phenotypes. We also show our set of phylogeny-constrained variant calls along with openly released sequencing data can be a valuable resource for the development of new computational methods.

Genetic regulation of splicing uniquely contributes to trait-associated genome-wide association studies (GWAS) signals. However, quantitative trait loci (QTL) analysis using short-read sequencing of bulk tissues fails to capture full-length and cell-type-specific isoforms. Here, we present an isoform-level lung cell atlas from 129 never-smoking Korean women using single-cell long-read RNA-sequencing, identifying abundant unannotated and cell-type-specific isoforms. Isoform-level signatures of 37 lung cell types display a larger difference and therefore improve cell-type classification compared to gene-level expression. Notably, isoform-QTLs (isoQTLs) detect unannotated and/or cell-type-specific isoforms with independent genetic regulation from expression-QTL (eQTL), supported by enriched splicing functional elements. IsoQTLs nominate susceptibility isoforms from previously unexplained lung function and cancer GWAS loci, via eQTL-independent signals. We highlight a potentially functional novel variant of PPIL6 in multiciliated cells underlying lung cancer risk through alternative splicing. This isoform-level resource advances our understanding of cell-type-specific isoform regulation and its contribution to lung traits and diseases.

Caseinolytic protease proteolytic subunit (ClpP) is part of the mitochondrial ClpXP protease responsible for degrading damaged proteins in the matrix, thus maintaining metabolic homeostasis. Small molecule activators of ClpP (ClpP agonists) have recently shown great promise against metabolically active cancers. TR107, a novel ClpP activator, demonstrates potent and selective cytotoxicity against glioma cells at nanomolar concentrations. Compared to the FDA-approved ClpP agonist ONC201 (Dordaviprone), TR107 shows greater efficacy across patient-derived and isogenic glioma models. TR107 effects are ClpP-dependent and result in widespread mitochondrial dysfunction evident by extensive protein degradation, impaired OxPhos, reduced mtDNA copy number, and ATP depletion. These disruptions are more severe in IDH-mutant than IDH-wildtype glioma cells, leading to enhanced cell cycle arrest, apoptosis, and inhibition of mTOR/AKT/4EBP1 signaling pathway. In this study, we show that ClpP agonism demonstrates preferential anti-tumor activity against IDH-mutant glioma in in vitro, ex vivo, and in vivo models. These findings support TR107 as a promising new targeted therapeutic for IDH-mutant gliomas.

Rapid and comprehensive analysis of complex proteomes across large sample sets is vital for unlocking the potential of systems biology. We present a high-throughput mass spectrometry (MS) proteomics method that integrates narrow-window data-independent acquisition (nDIA) with short-gradient micro-flow chromatography, enabling profiling of >240 samples per day. This optimized MS approach identifies 6,201 and 7,466 human proteins with 1- and 2-min gradients, respectively. As a practical application, we analyzed 507 samples composed of 13 different tissues from mice treated with the enzyme-drug L-asparaginase (ASNase) or its glutaminase-free Q59L mutant, generating a quantitative profile of 11,472 proteins following drug treatment. The MS results confirmed the impact of ASNase on amino acid metabolism in solid tissues. Further analysis revealed broad suppression of anticoagulants and cholesterol metabolism and uncovered numerous tissue-specific dysregulated pathways. In summary, the optimized high-throughput proteomics method accelerates systems-level analysis of a preclinical model to generate biological insights and clinically actionable hypotheses.

Abstract Most current large-scale whole-genome sequencing projects rely on short-read sequencing to call germline and somatic SVs, however it provides an incomplete view of the somatic variation landscape because of mappability limitations. In contrast, long-read sequencing can resolve highly repetitive regions of the human genome, and assemble variants into contiguous haplotypes, and therefore is a promising approach to resolve the hidden complexity of a cancer genome. However there is still a limited number of publicly available datasets for benchmarking and development of new methods. To motivate the development of new short- and long-read tools for cancer genomics, we created the CASTLE panel, based on multi-technology whole-genome sequencing of six commercially available tumor/normal cell line pairs (HCC1954, HCC1937, H1437, H2009, Hs578T and HCC1395). The panel represents two lung and three breast cancer cell lines. Genomic sequencing currently includes PacBio, Oxford Nanopore, Illumina, Hi-C and PoreC, in most cases sequenced from the same DNA extraction or cell line passage. We further generated high-confidence benchmarking somatic variant calls for SNPs, small indels and structural variants using the ensemble method. For structural variants, We used Severus, nanomonsv, SAVANA, Sniffles2, SvABA, GRIDSS, and Manta to generate initial variant calls; confident calls were defined if supported by at least two (out of three) technologies and at least 4 (out of 11) callers. For small variant benchmarking sets, we used a combination of Strelka2, DeepSomatic and ClairS. Overall, we release a new public resource for cancer genomic developments and benchmarking, which we are aiming to complement with additional genomic and transcriptomic technologies. The data and benchmarking datasets are openly available at: https://github.com/CASTLE-Panel/castle. Citation Format: Mikhail Kolmogorov, Ayse Gokce Kesus, Asher Bryant, Tanveer Ahmad, Byunggil Yoo, Sergey Aganezov, Anton Goretsky, Ataberk Donmez, Lisa Lansdon, Joshua Gardner, Brandy McNulty, Samuel Sacco, Jyoti Shetty, Yongmei Zhao, Bao Tran, Giuseppe Narzisi, Adrienne Hellend, Chengpeng Bi, Adam Walter, Margaret Gibson, Irina Pushel, Erin Guest, Tomi Pastinen, Nicolas Robine, Karen H. Miga, Midhat S. Farooqi, Benedict Paten. CASTLE: long-read sequencing panel of cancer cell lines to improve standards of somatic variant calling and benchmarking [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2026; Part 1 (Regular Abstracts); 2026 Apr 17-22; San Diego, CA. Philadelphia (PA): AACR; Cancer Res 2026;86(7 Suppl):Abstract nr 1510.

Somatic variant detection is an integral part of cancer genomics analysis. While most methods have focused on short-read sequencing, long-read technologies offer potential advantages in repeat mapping and variant phasing. We present DeepSomatic, a deep-learning method for detecting somatic small nucleotide variations and insertions and deletions from both short-read and long-read data. The method has modes for whole-genome and whole-exome sequencing and can run on tumor–normal, tumor-only and formalin-fixed paraffin-embedded samples. To train DeepSomatic and help address the dearth of publicly available training and benchmarking data for somatic variant detection, we generated and make openly available the Cancer Standards Long-read Evaluation (CASTLE) dataset of six matched tumor–normal cell line pairs whole-genome sequenced with Illumina, PacBio HiFi and Oxford Nanopore Technologies, along with benchmark variant sets. Across samples, both cell line and patient-derived, and across short-read and long-read sequencing technologies, DeepSomatic consistently outperforms existing callers. Somatic small variants in cancer genomes are identified in both short-read and long-read data.

Isocitrate dehydrogenase (IDH)-mutant gliomas have distinctive metabolic and biological traits that potentially render them susceptible to targeted treatments. Here, by conducting a high-throughput drug screen, we pinpointed a specific vulnerability of IDH-mutant gliomas to zotiraciclib (ZTR). ZTR exhibited selective growth inhibition across multiple IDH-mutant glioma in vitro and in vivo models. Mechanistically, ZTR at low doses suppressed CDK9 and RNA Pol II phosphorylation in IDH-mutant cells, disrupting mitochondrial function and NAD+ production, resulting in oxidative stress. Integrated biochemical profiling of ZTR kinase targets and transcriptomics unveiled that ZTR-induced bioenergetic failure was linked to the suppression of PIM kinase activity. We posit that the combination of mitochondrial dysfunction and an inability to adapt to oxidative stress resulted in significant cell death upon ZTR treatment, ultimately increasing the therapeutic vulnerability of IDH-mutant gliomas. These findings prompted a clinical trial evaluating ZTR in IDH-mutant gliomas (NCT05588141).

Cowpea mosaic virus (CPMV) is a plant virus with potent anti-tumor efficacy. Intratumoral CPMV, but not cowpea chlorotic mottle virus (CCMV), reprograms the tumor microenvironment priming durable, systemic anti-tumor immunity. Here, we performed a side-by-side mechanism of action comparison between CPMV and CCMV. Through immune-multiplexing and trafficking studies we observed that CPMV and CCMV are taken up by innate immune cells at similar rates but are processed differently. While CCMV induces proinflammatory interleukins 22, 23, and 27, CPMV is a strong inducer of type I, II, and III interferons, which contribute to the activation and enhancement of immune cell function. We observed that CPMV RNAs activate Toll-like receptor (TLR) 7 upon delivery—and not CCMV RNAs; CPMV RNAs persist longer within cells compared with CCMV. Taken together, this work identifies determinants for CPMV’s unique mechanism for intratumoral immunotherapy. It also broadens our understanding of the utility of plant viruses for biomedical applications.

Limitations to many current aqueous-based tryptic digestion methods include lengthy digestion times and both relatively high inter- and intra-day variability for both characteristic peptides identified and sequence coverages. This report describes results from digestion of some complex biomedical samples using the rapid Denaturing Organic Digestion method (DOD), an organic solvent-modified digestion method previously optimized for targeted protein digestion. Advantages of the DOD method included a very rapid digestion only requiring inexpensive solvents and reagents generally available in the laboratory, with no requirement for specialized equipment or expensive, specialized consumables. For this study, samples of E. coli and murine ileum protein extracts, and K562, a mass spectrometry-compatible human protein extract and reference standard routinely used to evaluate methods, were digested. Sequence coverage and characteristic peptide identification results were compared to those from 18 and 24 h conventional aqueous-based digestion methods. Across the samples tested, though the number of characteristic peptides and sequence coverages produced by the 5 min DOD method were very similar to those produced by the aqueous-based digestion methods, the specific characteristic proteins and their corresponding tryptic peptides identified following DOD method digestion included more hydrophilic and less hydrophobic species. In addition, we explored the effect of increasing digestion times with complex samples from 5 to 30 and 90 min for the DOD method. Increasing the digestion time to >= 30 min resulted in improved intra-day precision and the identification of many more peptide products than the currently used aqueous methods to which it was compared. These results suggest that the DOD organic-modified digestion method could, while markedly reducing protein digestion time, also provide more precise analysis and access to a somewhat different area of the proteome than that provided by current aqueous-based digestion methods. Significance: The DOD tryptic digest method is a very simple and rapid process with no requirement for expensive equipment or consumables. The method markedly reduces tryptic digestion time and cost, and substantially improves within-batch and across-analyst precision for peptide and sequence coverage results over methods to which it was compared. Importantly, it also provides access to a somewhat different subset of the proteome with different peptide products identified as compared to aqueous solvent-based digestion providing potential for increased proteome coverage for bottom-up analysis if used in conjunction with aqueous-based methods.

Introduction: Diffuse large B-cell lymphoma (DLBCL) is stratified into genetic subtypes (MCD, BN2, A53, N1, EZB, ST2) that differ in their gene expression profiles, oncogenic mechanisms, and response to therapy. Using paired single cell RNA (scRNA) and ATAC (scATAC) sequencing in DLBCL tumors, we previously identified gene expression themes reflecting B cell differentiation, cell growth, and cell cycle that distinguished intratumoral genetic subclones (Wang B, ASH, 2024). Here, we present a global analysis of transcription factor (TF) binding and activity in normal B cells and DLBCL tumors that revealed epigenetic heterogeneity among the DLBCL genetic subtypes, which underpins their divergent therapeutic responses. Methods: Paired scRNA and scATAC sequencing was performed on 102 DLBCL cases (504,444 cells) and 3 tonsils (12,227 cells). Gene expression and genetic subtypes were determined from matched bulk samples by RNA and whole exome sequencing. Computational analysis was performed using R/python and custom bioinformatic pipelines. Results: By linking TF activators (+/+) and repressors (-/+) to target gene expression using SCENIC+ (Bravo Gonzalez-Blas C, Nat Methods, 2023), we identified gene regulatory networks (GRNs) composed of enhancer-driven Regulons (eRegulons). In tonsillar B cell subpopulations, we identified 173 eRegulons that linked TF binding to 9,456 genomic regions and 3,535 target genes. A subset of these eRegulons were differentially active (p<0.05) in germinal center (GC) B cells (FOXO1, MEF2B, EBF1, PAX5, TCF3), plasma cells (PC; IRF4, XBP1, PRDM1) and memory B cells (KLF2, STAT1, ETV6). Importantly, cell lineage analysis traced the activity of these eRegulons along 3 differentiation trajectories stemming from naïve B cells towards either GC dark zone, PC, or memory B cells. Next, we used TF binding to define the epigenetic landscape of the DLBCL genetic subtypes, which could be distinguished from each other using subtype-specific gene expression signatures. Chromatin binding by TFs that regulate PC differentiation (IRF4, POU2F2, TCF4) correlated with the MCD, BN2 and A53 gene expression signatures as well as with gene expression themes reflecting PC differentiation, cell cycle, and cell growth. Binding by another group of TFs (FOXO1, MYBL1, STAT6, PAX5) was associated with the EZB signature and the GC B cell gene expression theme. Binding by BCL6 was anticorrelated with signatures of the N1 subtype and memory differentiation, suggesting that BCL6 antagonizes memory B cell differentiation and the generation of N1 DLBCL. To define GRNs in DLBCL genetic subtypes and genetic subclones, we used SCENIC+ to infer 289 eRegulons, comprised of 12,016 TF binding regions and 4,723 target genes. By integrating the eRegulon RNA and ATAC scores using multiomics factor analysis (Argelaguet R, Genome Biol, 2020), we identified major axes of variation that discriminated both DLBCL subtypes and normal B cell populations. The EZB and ST2 subtypes were significantly associated (p<0.05) with eRegulons that typify normal GC B cells (MEF2B, MEF2C, IRF8, FOXO1). Within these subtypes, subclones with REL amplification had significantly greater activity of a REL +/+ eRegulon than those with wild type REL (p<0.001). The MCD, A53 and BN2 subtypes were enriched (p<0.05) for the IRF4 +/+ eRegulon while MCD was additionally associated (p<0.05) with BATF, SPIB, XBP1 and PRDM1 eRegulons. A TBL1XR1 –/+ eRegulon was significantly associated with the N1 subtype (p<0.001), which is notable given that TBL1XR1 is a tumor suppressor that is frequently inactivated in N1. The subtype-associated eRegulons were also differentially active in normal B cell populations, with several MCD eRegulons active in PCs, N1 eRegulons active in memory B cells, and EZB eRegulons active in GC B cells. Accordingly, eRegulon scores correlated with the B cell differentiation themes across DLBCL subclones. Conclusions: By paired scRNA and scATAC sequencing, we identified GRNs present in normal and malignant B cells that highlight transcriptional states of DLBCL genetic subtypes which vary along three principal differentiation axes – GC B cell, memory B cell and PC. Our analysis illuminates the biological heterogeneity of DLBCL molecular subtypes and offers rationale targets for future therapeutic development.

An in vitro method for monitoring nanoparticle effects on IgE-dependent mast cell degranulation was developed and validated. The assayed nanoparticles included four clinical-grade nanomedicines (Abraxane, Doxil, AmBisome, and Feraheme) and three commercial research-grade nanomaterials (generation 5 PAMAM dendrimers with carboxy-, hydroxy-, or amine- surface functionalities). Most of the tested materials did not alter IgE-dependent mast cell degranulation, suggesting that nanoparticles and nanomedicines are unlikely to worsen pre-existing allergies to other antigens. Two clinical-grade formulations containing cytotoxic oncology drugs-Abraxane and Doxil-decreased degranulation. Abraxane but not Doxil decreased FcεR expression on the cell surface. Single-cell sequencing revealed the most differentially expressed genes (DEG) in Abraxane and Doxil-treated cultures. Interestingly, Feraheme and amine-terminated dendrimers induced DEG without affecting degranulation. These data demonstrate that some nanomaterials have more effects on immune cells than can be detected by a functional immunoassay.

Abstract Acute gastrointestinal intestinal GVHD (aGI-GVHD) is a serious complication of allogeneic hematopoietic stem cell transplantation, and the intestinal microbiota is known to impact on its severity. However, an association between treatment response of aGI-GVHD and the intestinal microbiota has not been well-studied. In a cohort of patients with aGI-GVHD (n=37), we found that non-response to standard therapy with corticosteroids was associated with prior treatment with carbapenem antibiotics and loss of Bacteroides ovatus from the microbiome. In a mouse model of carbapenem-aggravated GVHD, introducing Bacteroides ovatus reduced severity of GVHD and improved survival. Bacteroides ovatus reduced degradation of colonic mucus by another intestinal commensal, Bacteroides thetaiotaomicron, via its ability to metabolize dietary polysaccharides into monosaccharides, which then inhibit mucus degradation by Bacteroides thetaiotaomicron and reduce GVHD-related mortality.

Ion suppression is a major problem in mass spectrometry (MS)-based metabolomics; it can dramatically decrease measurement accuracy, precision, and signal-to-noise sensitivity. Here we report a new method, the IROA TruQuant Workflow, that uses a stable isotope-labeled internal standard (IROA-IS) plus novel companion algorithms to 1) measure and correct for ion suppression, and 2) perform Dual MSTUS normalization of MS metabolomic data. We have evaluated the method across ion chromatography (IC), hydrophilic interaction liquid chromatography (HILIC), and reverse phase liquid chromatography (RPLC)-MS systems in both positive and negative ionization modes, with clean and unclean ion sources, and across different biological matrices. Across the broad range of conditions tested, all detected metabolites exhibited ion suppression ranging from 1% to 90+% and coefficient of variations ranging from 1% to 20%, but the Workflow and companion algorithms were highly effective at nulling out that suppression and error. Overall, the Workflow corrects ion suppression across diverse analytical conditions and produces robust normalization of non-targeted metabolomic data.

Most current studies rely on short-read sequencing to detect somatic structural variation (SV) in cancer genomes. Long-read sequencing offers the advantage of better mappability and long-range phasing, which results in substantial improvements in germline SV detection. However, current long-read SV detection methods do not generalize well to the analysis of somatic SVs in tumor genomes with complex rearrangements, heterogeneity, and aneuploidy. Here, we present Severus: a method for the accurate detection of different types of somatic SVs using a phased breakpoint graph approach. To benchmark various short- and long-read SV detection methods, we sequenced five tumor/normal cell line pairs with Illumina, Nanopore, and PacBio sequencing platforms; on this benchmark Severus showed the highest F1 scores (harmonic mean of the precision and recall) as compared to long-read and short-read methods. We then applied Severus to three clinical cases of pediatric cancer, demonstrating concordance with known genetic findings as well as revealing clinically relevant cryptic rearrangements missed by standard genomic panels.

Hepatocellular carcinoma (HCC) is a molecularly heterogeneous solid malignancy, and its fitness may be shaped by how its tumor cells evolve. However, ability to monitor tumor cell evolution is hampered by the presence of numerous passenger mutations that do not provide any biological consequences. Here we develop a strategy to determine the tumor clonality of three independent HCC cohorts of 524 patients with diverse etiologies and race/ethnicity by utilizing somatic mutations in cancer driver genes. We identify two main types of tumor evolution, i.e., linear, and non-linear models where non-linear type could be further divided into classes, which we call shallow branching and deep branching. We find that linear evolving HCC is less aggressive than other types. GTF2IRD2B mutations are enriched in HCC with linear evolution, while TP53 mutations are the most frequent genetic alterations in HCC with non-linear models. Furthermore, we observe significant B cell enrichment in linear trees compared to non-linear trees suggesting the need for further research to uncover potential variations in immune cell types within genomically determined phylogeny types. These results hint at the possibility that tumor cells and their microenvironment may collectively influence the tumor evolution process. Clonality study in HCC finds diverse evolution patterns. Linear HCC is less aggressive, with GTF2IRD2B driver mutations. Non-linear has shallow/deep branching patterns with frequent TP53 driver mutations.

Somatic variant detection is an integral part of cancer genomics analysis. While most methods have focused on short-read sequencing, long-read technologies now offer potential advantages in terms of repeat mapping and variant phasing. We present DeepSomatic, a deep learning method for detecting somatic SNVs and insertions and deletions (indels) from both short-read and long-read data, with modes for whole-genome and exome sequencing, and able to run on tumor-normal, tumor-only, and with FFPE-prepared samples. To help address the dearth of publicly available training and benchmarking data for somatic variant detection, we generated and make openly available a dataset of five matched tumor-normal cell line pairs sequenced with Illumina, PacBio HiFi, and Oxford Nanopore Technologies, along with benchmark variant sets. Across samples and technologies (short-read and long-read), DeepSomatic consistently outperforms existing callers, particularly for indels.

Abstract BACKGROUND Targeting mitochondrial function has emerged as a promising therapeutic strategy in cancer treatment. TR107 is characterized as a specific activator of caseinolytic protease proteolytic subunit (ClpP) in the mitochondria, potentially affecting oncogenic pathways. Here, we investigated the effects of TR107 on adult malignant gliomas, IDH-wildtype and IDH-mutant METHODS We utilized patient-derived glioma cells to determine the efficacy of TR107 with cell viability, proliferation, cell cycle, and apoptosis assays. Immunofluorescent staining and electron microscopy (EM) were employed to uncover organelle morphological changes upon treatment. Mitochondrial function and ATP production were assessed in live cells utilizing Seahorse analysis. Global proteomics and RNA sequencing were performed to identify treatment-induced dysregulated pathways in glioma cells RESULTS Both IDH-wildtype and IDH-mutant glioma cell lines were highly sensitive to TR107, showing at least 100 times lower IC50 compared to the related compound ONC201. Genetic knock out of ClpP revealed that the growth inhibitory effects of TR107 are ClpP-dependent. Interestingly, IDH-mutant cell lines showed a more profound disintegration of mitochondrial morphology and dysregulation of mitochondrial function evidenced by an enhanced suppression of oxidative phosphorylation, mitochondrial complexes expression, mtDNA copy number, complex I activity, and ATP production, compared to IDH-wildtype cells. Six days of treatment with TR107 led to a complete inhibition of cell growth and up to 97% of cell death in IDH-mutant cells, while 50% of IDH-wildtype cells remained viable. Western blot, EM, and RNAsequencing analyses uncovered autophagy as a potential pro-survival mechanism in IDH-wildtype cells. Importantly, multiple metabolic and DNA repair-related pathways were affected by the treatment, suggested by global proteomics and RNA sequencing. Finally, TR107 was not affected by ABC transporters, supporting its potential use in brain tumor patients CONCLUSION TR107 demonstrated potent cytotoxic effects in patient-derived glioma cells by disrupting mitochondrial structural and functional integrity. TR107 efficacy in vivo is under investigation.

Loss of BRCA2 (breast cancer 2) is lethal for normal cells. Yet it remains poorly understood how, in BRCA2 mutation carriers, cells undergoing loss of heterozygosity overcome the lethality and undergo tissue-specific neoplastic transformation. Here, we identified mismatch repair gene mutL homolog 1 (MLH1) as a genetic interactor of BRCA2 whose overexpression supports the viability of Brca2-null cells. Mechanistically, we showed that MLH1 interacts with Flap endonuclease 1 (FEN1) and competes to process the RNA flaps of Okazaki fragments. Together, they restrained the DNA2 nuclease activity on the reversed forks of lagging strands, leading to replication fork (RF) stability in BRCA2-deficient cells. In these cells, MLH1 also attenuated R-loops, allowing the progression of stable RFs, which suppressed genomic instability and supported cell viability. We demonstrated the significance of their genetic interaction by the lethality of Brca2-mutant mice and inhibition of Brca2-deficient tumor growth in mice by Mlh1 loss. Furthermore, we described estrogen as inducing MLH1 expression through estrogen receptor α (ERα), which might explain why the majority of BRCA2 mutation carriers develop ER-positive breast cancer. Taken together, our findings reveal a role of MLH1 in relieving replicative stress and show how it may contribute to the establishment of BRCA2-deficient breast tumors.

A tumor ecosystem constantly evolves over time in the face of immune predation or therapeutic intervention, resulting in treatment failure and tumor progression. Here, we present a single-cell transcriptome-based strategy to determine the evolution of longitudinal tumor biopsies from liver cancer patients by measuring cellular lineage and ecology. We construct a lineage and ecological score as joint dynamics of tumor cells and their microenvironments. Tumors may be classified into four main states in the lineage-ecological space, which are associated with clinical outcomes. Analysis of longitudinal samples reveals the evolutionary trajectory of tumors in response to treatment. We validate the lineage-ecology-based scoring system in predicting clinical outcomes using bulk transcriptomic data of additional cohorts of 716 liver cancer patients. Our study provides a framework for monitoring tumor evolution in response to therapeutic intervention.