Biology subjects

Beder, T.

Publications and source records attributed to Beder, T..

7 recordsLinked to original sources

bioRxiv2026

ALLCatchR, a machine-learning classifier identifies now 20 T-ALL subtypes across cohorts and age groups

T-cell acute lymphoblastic leukemia (T-ALL) comprises molecularly diverse subtypes, but robust cross-cohort validations and operational gene-expression definitions are lacking. To establish a gene-expression-anchored framework for T-ALL subtyping, we aggregated 2,314 transcriptomes (15 cohorts, age: 0.8 to 90.8 years). An extended unsupervised approach defined 17 main clusters and 3 subclusters in samples with high blast fractions. Supervised analyses added an overarching immature T-ALL (ETP-like) definition and resolved the LMO2{gamma}{delta} -like subtype. All clusters contained samples from at least two cohorts. Characteristic genomic driver enrichments were consistent across cohorts, while gene expression clusters did not correspond exclusively to single driver events but also reflected developmental origins. A machine learning classifier based on ALLCatchR, our B-ALL classifier, identified these 20 transcriptomic subtypes and the immature T-ALL (ETP-like) signature with 0.995-1.0 accuracy in a validation set (n=203). Testing the classifier on a second hold-out data set (n=265 samples) showed that 92.7% of predictions matched with corresponding driver alterations. Across all samples, 83.2% of cases received high-confidence predictions, 7.3% candidate predictions, and 9.5% remained unclassified, largely because of low blast fractions. We identified a novel gene expression cluster markedly enriched (P<0.001) for clonal hematopoiesis mutations (IDH2 R140Q, DNMT3A) and a stem-/progenitor cell-like gene expression. This novel clonal hematopoiesis-related T-ALL subtype was observed in six cohorts representing 8.9% of adults and 39.5% of patients aged >50 years. We advanced ALLCatchR, as a free R package that now enables B-/T-lineage separation, gene-expression subtyping, blast estimation, and developmental annotation to harmonize T-ALL classification across studies and clinical contexts. Key PointsO_LIEstablished using 2,314 T-ALL transcriptomes, ALLCatchR2 assigns 20 RNA-Seq subtypes and their developmental underpinnings across ages. C_LIO_LIClonal hematopoiesis-related T-ALL (DNMT3A, IDH2 R140Q) defines an immature gene-expression cluster in [~]40% of patients aged >50 years. C_LI

bioinformatics↗

bioRxiv2025

IntegrateALL: an end-to-end RNA-seq analysis pipeline for multilevel data extraction and interpretable subtype classification in B-precursor ALL

Transcriptome sequencing (RNA-seq) is emerging as a diagnostic standard for B-cell precursor acute lymphoblastic leukemia (B-ALL). Expression-based classifiers reach [~]95% accuracy, but reproducible end-to-end solutions that also integrate transcript-derived genomic drivers and quantitative virtual karyotyping are lacking. We developed IntegrateALL, a Snakemake pipeline that standardizes RNA-seq analysis from FASTQ to rule-based subtype assignment across 26 WHO-HAEM5/ICC entities by integrating expression-based subtype prediction, gene fusion- / hotspot SNV calling and virtual karyotyping. We introduce KaryALL, a machine-learning classifier that uses normalized expression and minor-allele-frequency features (RNASeqCNV) to distinguish near haploid, hypodiploid and high hyperdiploid B-ALL and chromosome-21 gains/iAMP21 (accuracy: 0.98 / F1-score: 0.96 on 615 independent test samples). SNP-array concordance supported RNA-based karyotyping. Applied to 774 unselected B-ALL cases, IntegrateALL yielded unambiguous subtype assignments in 81.5%, based on concordance of gene expression class with a defining driver (75.3% of all cases) or, in selected cases, high-confidence expression-based classification alone (6.2%); the remainder (18.5%) were flagged for manual curation. Independent validation (3 cohorts; n=436, including pediatric cases) reproduced these distributions. Across all patients (n=1,210), 2.6% harbored two subtype defining drivers, including hyperdiploidy in fusion-driven subtypes where it was not expected or subtype-defining SNVs (e.g., PAX5 P80R / IKZF1 N159Y) co-occurring with BCR::ABL1-positive/-like, KMT2A- or DUX4-fusions. In most dual-driver cases, one subtype gene expression signature predominated, indicating a hierarchy of oncogenic control and the value of systematic driver screening alongside expression-based calls. IntegrateALL provides an adaptable fully reproducible workflow for molecular B-ALL characterization by systematically integrating genomic drivers and downstream gene regulation.

cancer biology↗

bioRxiv2025

Intra-subtype heterogeneity shapes treatment response in KMT2A-rearranged ALL across all age groups

BackgroundKMT2A-rearranged B-cell acute lymphoblastic leukemia (KMT2Ar B-ALL) exhibits significant heterogeneity in age of onset, developmental origins, and clinical outcomes. The interplay of individual factors influencing early treatment response within this high-risk molecular subtype remains poorly elucidated. We aimed to comprehensively assess how leukemic developmental state, fusion partner, and patient age jointly influence early treatment response and drug sensitivity. MethodsTo identify determinants of early treatment response to induction chemotherapy, we analysed 465 KMT2Ar B-ALL cases spanning a wide age range (1 month to 89 years) by integrating transcriptomic and genomic profiling with functional drug response and measurable residual disease (MRD) kinetics. Transcriptomic profiling was used to derive a developmental maturity score based on proximity to physiological B-cell differentiation from a normal B-lymphopoiesis reference. Using an ordinal regression model we identify gene expression programs associated with early MRD response. ResultsWe observed a strong inverse correlation between MRD clearance with advancing age (p=2.1E-04), proximity to early B-cell-precursor developmental state (low maturity score, p=1.3E-03) and AFF1 as fusion partner (p=7.0E-04). A combined model confirmed the predominate impact of both maturity and KMT2A fusion partner on MRD response, supporting the concept that the cells developmental state defines therapy response. Gene expression analysis identified cellular traits that relate to MRD response (e.g. chromatin organization, immune modulation and proliferation). This gene expression classifier grouped cases by MRD response but also by ex-vivo induction drug sensitivity. Notably, good responders to ex-vivo induction drugs were characterized by a higher maturity score (p=1.8E-03), whereas for less mature KMT2Ar B-ALL cases response profiles suggested higher Venetoclax sensitivity. ConclusionsOur study provides an integrative framework linking developmental phenotype, fusion partner, and MRD kinetics across the full age spectrum of KMT2Ar B-ALL. The maturity score, derived from bulk transcriptome data, offers a biologically relevant predictor of early treatment response and drug sensitivity. These insights may support future risk-adapted strategies and therapeutic targeting, particularly in immature KMT2Ar B-ALL.

cancer biology↗

bioRxiv2023

Mutational and transcriptional landscape of pediatric B-cell precursor lymphoblastic lymphoma

Pediatric B-cell precursor (BCP) lymphoblastic malignancies are neoplasms with manifestation either in bone marrow/blood (BCP acute lymphoblastic leukemia, BCP-ALL) or less common in extramedullary tissue (BCP lymphoblastic lymphoma, BCP-LBL). Although both presentations are similar in morphology and immunophenotype molecular studies are virtually restricted to BCP-ALL so far. The lack of molecular studies on BCP-LBL is probably due to its rarity and the restriction to tiny, mostly formalin-fixed paraffin embedded (FFPE) tissues. Here we present the first comprehensive mutational and transcriptional analysis of what we consider the largest BCP-LBL cohort described to date (n=97). Whole exome sequencing indicates a mutational spectrum of BCP-LBL strikingly similar to that found in BCP-ALL. However, epigenetic modifiers were more frequently mutated in BCP-LBL, whereas BCP-ALL was more frequently affected by mutation in genes involved in B-cell development. Integrating copy number alterations, somatic mutations and gene expression by RNA-sequencing revealed virtually all molecular subtypes originally defined in BCP-ALL to be present in BCP-LBL too, with only 7% of lymphomas that were not assigned to a subtype. Therefore, the results here described may pave the way for molecular risk adapted treatment protocols for BCP-LBL patients. KeypointsComprehensive molecular characterization of B-cell precursor lymphoblastic lymphoma allows molecular subtyping analogous to leukemias Compared to leukemias, lymphomas show more alterations in epigenetic modifiers and less in B-cell development genes

cancer biology↗

bioRxiv2023

Machine learning on large scale perturbation screens for SARS-CoV-2 host factors identifies β-catenin/CBP inhibitor PRI-724 as a potent antiviral

Expanding antiviral treatment options against SARS-CoV-2 remains crucial as the virus evolves rapidly and drug resistant strains have emerged. Broad spectrum host-directed antivirals (HDA) are promising therapeutic options, however the robust identification of relevant host factors by CRISPR/Cas9 or RNA interference screens remains challenging due to low consistency in the resulting hits. To address this issue, we employed machine learning based on experimental data from knockout screens and a drug screen. As gold standard, we assembled perturbed genes reducing virus replication or protecting the host cells. The machines based their predictions on features describing cellular localization, protein domains, annotated gene sets from Gene Ontology, gene and protein sequences, and experimental data from proteomics, phospho-proteomics, protein interaction and transcriptomic profiles of SARS-CoV-2 infected cells. The models reached a remarkable performance with a balanced accuracy of 0.82 (knockout based classifier) and 0.71 (drugs screen based classifier), suggesting patterns of intrinsic data consistency. The predicted host dependency factors were enriched in sets of genes particularly coding for development, morphogenesis, and neural related processes. Focusing on development and morphogenesis-associated gene sets, we found {beta}-catenin to be central and selected PRI-724, a canonical {beta}-catenin/CBP disruptor, as a potential HDA. PRI-724 limited infection with SARS-CoV-2 variants, SARS-CoV-1, MERS-CoV and IAV in different cell line models. We detected a concentration-dependent reduction in CPE development, viral RNA replication, and infectious virus production in SARS-CoV-2 and SARS-CoV-1-infected cells. Independent of virus infection, PRI-724 treatment caused cell cycle deregulation which substantiates its potential as a broad spectrum antiviral. Our proposed machine learning concept may support focusing and accelerating the discovery of host dependency factors and the design of antiviral therapies. Authors summaryDrug resistance to pathogens is a well-known phenomenon which was also observed for SARS-CoV-2. Given the gradually increasing evolutionary pressure on the virus by herd immunity, we attempted to enlarge the available antiviral repertoire by focusing on host proteins that are usurped by viruses. The identification of such proteins was followed within several high throughput screens in which genes are knocked out individually. But, so far, these efforts led to very different results. Machine learning helps to identify common patterns and normalizes independent studies to their individual designs. With such an approach, we identified genes that are indispensable during embryonic development, i.e., when cells are programmed for their specific destiny. Shortlisting the hits revealed {beta}-catenin, a central player during development, and PRI-724, which inhibits the interaction of {beta}-catenin with cAMP responsive element binding (CREB) binding protein (CBP). In our work, we confirmed that the disruption of this interaction impedes virus replication and production. In A549-AT cells treated with PRI-724, we observed cell cycle deregulation which might contribute to the inhibition of virus infection, however the exact underlying mechanisms needs further investigation.

microbiology↗

bioRxiv2023

The gene expression classifier ALLCatchR identifies B-precursor ALL subtypes and underlying developmental trajectories across age

Current classifications (WHO-HAEM5 / ICC) define up to 26 molecular B-cell precursor acute lymphoblastic leukemia (BCP-ALL) disease subtypes, which are defined by genomic driver aberrations and corresponding gene expression signatures. Identification of driver aberrations by RNA-Seq is well established, while systematic approaches for gene expression analysis are less advanced. Therefore, we developed ALLCatchR, a machine learning based classifier using RNA-Seq expression data to allocate BCP-ALL samples to 21 defined molecular subtypes. Trained on n=1,869 transcriptome profiles with established subtype definitions (4 cohorts; 55% pediatric / 45% adult), ALLCatchR allowed subtype allocation in 3 independent hold-out cohorts (n=1,018; 75% pediatric / 25% adult) with 95.7% accuracy (averaged sensitivity across subtypes: 91.1% / specificity: 99.8%). High confidence predictions were achieved in 84.6% of samples with 99.7% accuracy. Only 1.2% of samples remained unclassified. ALLCatchR outperformed existing tools and identified novel candidates in previously unassigned samples. We established a novel RNA-Seq reference of human B-lymphopoiesis. Implementation in ALLCatchR enabled projection of BCP-ALL samples to this trajectory, which identified shared patterns of proximity of BCP-ALL subtypes to normal lymphopoiesis stages. ALLCatchR sustains RNA-Seq routine application in BCP-ALL diagnostics with systematic gene expression analysis for accurate subtype allocations and novel insights into underlying developmental trajectories.

cancer biology↗

bioRxiv2021

Identifying essential genes across eukaryotes by machine learning

Identifying essential genes on a genome scale is resource intensive and has been performed for only a few eukaryotes. For less studied organisms essentiality might be predicted by gene homology. However, this approach cannot be applied to non-conserved genes. Additionally, divergent essentiality information is obtained from studying single cells or whole, multi-cellular organisms, and particularly when derived from human cell line screens and human population studies. We employed machine learning across six model eukaryotes and 60,381 genes, using 41,635 features derived from sequence, gene functions and network topology. Within a leave-one-organism-out cross-validation, the classifiers showed a high generalizability with an average accuracy close to 80% in the left-out species. As a case study, we applied the method to Tribolium castaneum and validated predictions experimentally yielding similar performance. Finally, using the classifier based on the studied model organisms enabled linking the essentiality information of human cell line screens and population studies.

bioinformatics↗