MTMartin S. Taylor

Martin S. Taylor

*This is an algorithmically generated author record based on publicly available information.

Current Organizations / Institutes

The organizations/institutes from the scholar's latest article

Previous Organizations / Institutes

The organizations / institutes from the scholar's other articles in the past ten years


Martin S. Taylor has a total of 988 co-authors. Here are the top co-authors by number of shared publications:


Latest Publications

  • Open Access

    Explainable machine learning relates histological to genomic pathology

    by John Connelly, Barbara Hernando, Juliet Luft, Craig J. Anderson, Peter Bankhead, Frances Connor, Stuart Aitken, Liver Cancer Evolution Consortium, Colin A. Semple, Paul Flicek, et al.

    Background & Aims: Haematoxylin and eosin (H&E) staining remains the diagnostic gold standard for solid cancers, including hepatocellular carcinoma, and is increasingly complemented by genomic profiling for precision medicine. Inferring genomic alterations directly from H&E images could streamline testing, but heterogeneity and biases in human training data limit interpretation of genotype-phenotype associations. Here, we aimed to relate histologic to genomic pathology to provide biological explainability for mutation prediction models and assess the impact of germline variation on model performance. Methods: We analysed 597 murine liver tumours with matched whole-genome sequencing and histopathology (163,835 image tiles; 22.9 million nuclei). Our controlled in vivo design accounted for germline variation, biological sex, and causal mutagen (N-diethylnitrosamine), removing confounding factors present in human cohorts. We trained and evaluated deep learning and supervised machine learning models to predict germline variation and cancer driver alterations from H&E. Results: Modelling accurately predicted germline and somatic alterations from histology, at both locus-specific and genome-wide scales. Quantitative image analysis revealed an unexpected association between Egfr driver mutations and hepatic steatosis, linking genotype to an interpretable morphological phenotype. While model performance declined when applied to tumours from unrepresented genetic backgrounds, this limitation was biologically informative, revealing strain-dependent differences in tumour evolution, notably the prevalence of whole-genome duplication. Conclusions: Machine learning integration of histological and genomic pathology enables accurate, interpretable inference of genetic alterations from H&E, potentially reducing reliance on costly ancillary molecular assays. Our predictions are supported by human-interpretable biological features, addressing concerns around “black-box” technologies. However, caution is required when applying such methods to samples with a genetic background that, even if closely related, is beyond the genetic horizon of training data.

  • Open Access

    Genetic background sets the trajectory of experimental cancer evolution

    by Sarah J. Aitken, Frances Connor, Christine Feig, Tim F. Rayner, Margus Lukk, Juliet Luft, Stuart Aitken, Claudia Arnedo-Pac, James F. Hayes, Michael D. Nicholson, et al.

    Human cancers are heterogeneous 1 . Dissecting how germline genetic variation and environmental factors shape tumour evolution using human datasets is limited by inherent diversity in genetic backgrounds 2 and environmental exposures 3–5 . Here, to overcome these limitations, we re-ran early tumour evolution hundreds of times in diverged inbred mouse strains, generating matched histology and whole-genome and transcriptome sequences. The sex, environment and carcinogenic exposures were all controlled, and the study design allowed us to capture genetic variation comparable with that observed across human populations while exploiting the nested hierarchical structure of strain–litter–animal–tumour relationships. Our analyses reveal that epistatic interactions between genetic background and acquired somatic mutations result in population-specific disease progression, including choice of driver mutations, occurrence of whole-genome duplication and subclonal selection dynamics that mirror both cancer susceptibility and tumour growth rate. Even modest genetic divergence, comparable with that found across human ancestry groups, can strikingly alter selection pressures during cancer development to shape both cancer risk and the trajectory of tumour evolution.


  • Open Access

    Figure 4 from In Vivo Modeling of Patient Genetic Heterogeneity Identifies New Ways to Target Cholangiocarcinoma

    by Nicholas T. Younger, Mollie L. Wilson, Anabel Martinez Lyons, Edward J. Jarman, Alison M. Meynert, Graeme R. Grimes, Konstantinos Gournopanos, Scott H. Waddell, Peter A. Tennant, David H. Wilson, et al.

    Therapeutic coinhibition of Wnt and PI3K signaling reduces tumor growth in ICC. A, RNA-seq data of human ICC demonstrating a positive correlation between the activity of canonical Wnt signaling and Akt signaling. B, Schematic representation of the KPPTom cholangiocarcinoma model where CreERT expression in Keratin-19–positive cholangiocytes results in the inactivation of Trp53 and Pten, whereas labeling transformed cells with tdTomato. C, Representative IHC staining of KPPTom model following tamoxifen administration (day 0) and following 4 and 8 weeks of thioacetamide administration. tdTomato (red) denotes recombined cholangiocytes (denoted by Keratin-19; green). Blue, DNA. Top, whole mount FUNGI images; bottom, 2D histologic sections. Scale bar, 200 μm. White arrows, tdTomato-positive cells. D, Quantification of liver tissue occupied by tumor in the KPPTom ICC model. E, IHC showing that KPPTom ICC has activated canonical Wnt signaling [by staining for dephosphorylated (active) β-catenin] and PI3K activity (through pAKTSer647 positivity). Red arrows, positive cells. Scale bar, 100 μm. F, A schematic representation of how the KPPTom model was used to test the effectiveness of Wnt and Pi3K inhibitor combinations on ICC progression. G, IHC staining for tdTomato-positive cancer cells in vehicle-treated animals compared with those treated with a combination of LGK974 and pictilisib. Scale bar, 100 μm. H, Number of tdTomato-positive cells in KPPTom animals given vehicle or LGK974 and pictilisib in combination. I, Proportion of KPPTom animals containing macroscopic tumors in KPPTom animals treated with vehicle versus combination treatment. b.d., bile duct; p.v., portal vein.

  • Open Access

    Figure 3 from In Vivo Modeling of Patient Genetic Heterogeneity Identifies New Ways to Target Cholangiocarcinoma

    by Nicholas T. Younger, Mollie L. Wilson, Anabel Martinez Lyons, Edward J. Jarman, Alison M. Meynert, Graeme R. Grimes, Konstantinos Gournopanos, Scott H. Waddell, Peter A. Tennant, David H. Wilson, et al.

    Nf2 loss results in RasG12D-induced oncogenesis and cooperates with Trp53 loss to accelerate ICC formation. A, Kaplan–Meier curve demonstrating the relative survival proportions of mice with KRASG12D and gRNAs targeting Trp53 (N = 12), Nf2 (N = 5), Nf2;Trp53 (N = 13), or nontargeting control (scrm, N = 5). B and C, Proportion of liver occupied by tumor (B) and number of tumors per mouse (C). D, Hematoxylin and eosin (H&E) staining of KRASG12D tumors with Trp53, Nf2, or Trp53;Nf2 loss. Scale bar, 100 μm. Dotted line, tumor-stroma boundary. E, Comparison of RNA-seq analysis when the transcriptomes from Nf2;Trp53 versus Trp53 alone tumors (blue) are compared with transcripts from Nf2;Trp53 versus Nf2 alone (yellow) tumors. Each group contains N = 4 regionally distinct tumors. F, Analysis of RPPA data demonstrating the changes in the proportion of phosphorylated GSK3α/β, β-catenin, and pAKT relative to total protein levels in KRASG12D;Trp53KO (gray points), KRASG12D;Nf2KO (yellow points), KRASG12D;Trp53KO;Nf2KO (blue points). G, IHC of active, dephosphorylated β-catenin (top) and phosphorylated AKTSer647 (bottom) in KRASG12D;Trp53KO, KRASG12D;Nf2KO, KRASG12D;Trp53KO/Nf2KO tumors. Scale bar, 50 μm. H. Immunoblot for dephosphorylated (active) β-catenin (β-cateninSer33/37/Thr41) in tumors isolated from mice baring KrasG12D -driven ICC with Trp53, Nf2, or Trp53;Nf2 co-loss. GAPDH was used as a loading control. I, Schematic representing our dosing approach to determine whether Wnt inhibition, PI3K inhibition, or a combination of the two is effective in improving the survival of mice with KRASG12D;Trp53KO;Nf2KO ICC. J, Kaplan–Meier curve demonstrating the survival changes when KRASG12D;Trp53KO;Nf2KO animals are treated with vehicle (yellow line), LGK974 (Wnt-inhibitor; blue line), pictilisib (PI3K inhibitor; orange line), or a combination (green line; N = 5 per group).

  • Open Access

    Figure 1 from In Vivo Modeling of Patient Genetic Heterogeneity Identifies New Ways to Target Cholangiocarcinoma

    by Nicholas T. Younger, Mollie L. Wilson, Anabel Martinez Lyons, Edward J. Jarman, Alison M. Meynert, Graeme R. Grimes, Konstantinos Gournopanos, Scott H. Waddell, Peter A. Tennant, David H. Wilson, et al.

    In silico screening identifies novel drivers of ICC. A, The number of samples with mutations in driver genes identified following analysis with IntOGen. Samples clustered into those that had been previously identified in ICC or are present in the COSMIC database (yellow points), or those mutated genes that have not previously been assigned as being cancer drivers (blue points). B, Frequency of samples containing known ICC or COSMIC mutation in each of the individual cohorts collated in this study and in the pooled datasets. C, Frequency of samples containing a predicted novel oncogenic mutation identified by IntOGen. In both B and C, heatmaps represent the frequency at which each mutation is found within each study and the aggregated frequency between all studies. Top bar represents the number of times a mutation was identified between studies. D, Pathway interaction analysis of putative ICC driver mutations identified by IntOGen based on known functional (solid lines) and predicted physical (dotted lines) interactions. Numbers 1 to 6 represent distinct gene relationship modules based on predicted or known genetic interactions. E, Co-occurrence and mutual exclusivity analysis demonstrates that there is a high level of mutational exclusivity between canonical driver mutations in patients with ICC. F, Co-occurrence and mutual exclusivity of COSMIC drivers that co-occur ≥15 times in the human sample set and combined “other,” which includes all novel drivers. N = 277 patient exomes or genomes with matched, noncancerous tissue.

  • Open Access

    Figure 2 from In Vivo Modeling of Patient Genetic Heterogeneity Identifies New Ways to Target Cholangiocarcinoma

    by Nicholas T. Younger, Mollie L. Wilson, Anabel Martinez Lyons, Edward J. Jarman, Alison M. Meynert, Graeme R. Grimes, Konstantinos Gournopanos, Scott H. Waddell, Peter A. Tennant, David H. Wilson, et al.

    In vivo CRISPR-Cas9 screening identifies transforming mutations that interact with mutant Ras. A, Schematic of this study in which high-content sequencing data are collated from patients with ICC and the mutational profile of these tumors rationalized to identify novel, high confidence drivers of ICC. These putative drivers are used as input for an in vivo SpCas9/CRISPR screen to identify novel functional processes that drive ICC growth. B, Macroscopic images of the livers following injection with either NRASG12V or KRASG12D alone (left) or in combination with ICCLib (right; dotted line, tumor). Scale bar, 1 cm. C, Quantification of macroscopic tumors per mouse at 10 weeks in mice bearing NrasG12V -expressing tumors and 8 weeks in those with KrasG12D-driven cancer. Each circle represents a different animal. D, The number of samples containing Indels in a particular gene following whole exome sequencing. Top graph lists those mutations found in both NRASG12V and KRASG12D tumors, and bottom graphs denote those mutations that are found only in KRASG12D - or NRASG12V -expressing tumors. Sample frequency (%) denotes the proportion of tumors containing any given mutation, whereas (count) is absolute number. (N represents anatomically discreet tumors recovered from at least four individual animals, KRASG12DN = 14 and NRASG12VN = 10.) E, PCA showing how samples group based on their transcriptomic signature and gRNA-induced mutations associated with each tumor type.

  • Open Access

    Data from In Vivo Modeling of Patient Genetic Heterogeneity Identifies New Ways to Target Cholangiocarcinoma

    by Nicholas T. Younger, Mollie L. Wilson, Anabel Martinez Lyons, Edward J. Jarman, Alison M. Meynert, Graeme R. Grimes, Konstantinos Gournopanos, Scott H. Waddell, Peter A. Tennant, David H. Wilson, et al.

    Intrahepatic cholangiocarcinoma (ICC) is an aggressive malignancy of the bile ducts within the liver characterized by high levels of genetic heterogeneity. In the context of such genetic variability, determining which oncogenic mutations drive ICC growth has been difficult, and developing modes of patient stratification and targeted therapies remains challenging. Here we model the interactions between rare mutations with more common driver genes and combine in silico analysis of patient data with highly multiplexed in vivo CRISPR-spCas9 screens to perform a functional in vivo study into the role genetic heterogeneity plays in driving ICC. Novel tumor suppressors were uncovered, which, when lost, cooperate with the RAS oncoprotein to drive ICC growth. Focusing on a set of driver mutations that interact with KRAS to initiate aggressive, sarcomatoid-type ICC revealed that tumor growth relies on Wnt and PI3K signaling. Pharmacologic coinhibition of Wnt and PI3K in vivo impeded ICC growth regardless of mutational profile. Therefore, Wnt and PI3K activity should be considered as a signature by which patients can be stratified for treatment independent of tumor genotype, and inhibitors of these pathways should be levied to treat ICC.Significance:This work shows that, despite significant genetic heterogeneity, intrahepatic cholangiocarcinoma relies on a limited number of signaling pathways to grow, suggesting common therapeutic vulnerabilities across patients.

  • Open Access

    Germline and somatic variation influence nuclear-mitochondrial crosstalk in tumourigenesis

    by Maëlle Daunesse, Vasavi Sundaram, Elissavet Kentepozidou, John Connelly, Craig J. Anderson, Liver Cancer Evolution Consortium, Núria López-Bigas, Colin A. Semple, Duncan T. Odom, Martin S. Taylor, et al.

    Hepatocellular carcinoma is the most prevalent primary liver cancer and arises from hepatocytes, which are very metabolically active and contain abundant mitochondria. Despite extensive characterisation of nuclear driver genes and pathways, the role of crosstalk between the mitochondrial and nuclear genomes in mitochondrial mutagenesis remains poorly understood. To address this, we leveraged a chemical carcinogenesis model of liver cancer in four mouse strains. We analysed hundreds of tumours with paired whole-genome sequencing and total RNA sequencing to comprehensively characterise mitochondrial mutations and expression. Following de novo assembly of strain-specific mitochondrial genomes, we devised an accurate and efficient heteroplasmy detection approach and analysed mitochondrial DNA (mtDNA) content and transcription. Across strains, there was high concordance of the number of heteroplasmies, mutation signatures, and variant allele frequencies. In contrast, the heteroplasmy rate and genic locations differed between strains, suggesting strain-specificity in DNA repair mechanisms in early tumour development. Independent of age or environmental factors, tumours had lower mtDNA content than adjacent normal liver, suggesting mitochondrial loss during tumourigenesis. Additionally, within individual strains, mtDNA content differences were associated with nuclear driver gene choice. Our comprehensive characterisation of mitochondrial genomes and expression patterns demonstrates that both germline and somatic genetic variation influence tumour mtDNA content, mutation patterns, and gene expression.


  • Open Access

    Genetic background sets the trajectory of cancer evolution

    by Sarah J. Aitken, Frances Connor, Christine Feig, Tim F. Rayner, Margus Lukk, Juliet Luft, Stuart Aitken, Claudia Arnedo-Pac, James F. Hayes, Michael D. Nicholson, et al.

    Human cancers are heterogeneous. Their genomes evolve from genetically diverse germlines in complex and dynamic environments, including exposure to potential carcinogens. This heterogeneity of humans, our environmental exposures, and subsequent tumours makes it challenging to understand the extent to which cancer evolution is predictable. Addressing this limitation, we re-ran early tumour evolution hundreds of times in diverse, inbred mouse strains, capturing genetic variation comparable to and beyond that found in human populations. The sex, environment, and carcinogenic exposures were all controlled and tumours comprehensively profiled with whole genome and transcriptome sequencing. Within a strain, there was a high degree of consistency in the mutational landscape, a limited range of driver mutations, and all strains converged on the acquisition of a MAPK activating mutation with similar transcriptional disruption of that pathway. Despite these similarities in the phenotypic state of tumours, different strains took markedly divergent paths to reach that state. This included pronounced biases in the precise driver mutations, the strain specific occurrence of whole genome duplication, and differences in subclonal selection that reflected both cancer susceptibility and tumour growth rate. These results show that interactions between the germline genome and the environment are highly deterministic for the trajectory of tumour genome evolution, and even modest genetic divergence can substantially alter selection pressures during cancer development, influencing both cancer risk and the biology of the tumour that develops.


  • Open Access

    Strand-resolved mutagenicity of DNA damage and repair

    by Craig J. Anderson, Lana Talmane, Juliet Luft, John Connelly, Michael D. Nicholson, Jan C. Verburg, Oriol Pich, Susan Campbell, Marco Giaisi, Pei-Chi Wei, et al.

    DNA base damage is a major source of oncogenic mutations1. Such damage can produce strand-phased mutation patterns and multiallelic variation through the process of lesion segregation2. Here we exploited these properties to reveal how strand-asymmetric processes, such as replication and transcription, shape DNA damage and repair. Despite distinct mechanisms of leading and lagging strand replication3,4, we observe identical fidelity and damage tolerance for both strands. For small alkylation adducts of DNA, our results support a model in which the same translesion polymerase is recruited on-the-fly to both replication strands, starkly contrasting the strand asymmetric tolerance of bulky UV-induced adducts5. The accumulation of multiple distinct mutations at the site of persistent lesions provides the means to quantify the relative efficiency of repair processes genome wide and at single-base resolution. At multiple scales, we show DNA damage-induced mutations are largely shaped by the influence of DNA accessibility on repair efficiency, rather than gradients of DNA damage. Finally, we reveal specific genomic conditions that can actively drive oncogenic mutagenesis by corrupting the fidelity of nucleotide excision repair. These results provide insight into how strand-asymmetric mechanisms underlie the formation, tolerance and repair of DNA damage, thereby shaping cancer genome evolution.