MFmap: A semi-supervised generative model matching cell lines to tumours and cancer subtypes

Translating in vitro results from experiments with cancer cell lines to clinical applications requires the selection of appropriate cell line models. Here we present MFmap (model fidelity map), a machine learning model to simultaneously predict the cancer subtype of a cell line and its similarity to an individual tumour sample. The MFmap is a semi-supervised generative model, which compresses high dimensional gene expression, copy number variation and mutation data into cancer subtype informed low dimensional latent representations. The accuracy (test set F1 score >90%) of the MFmap subtype prediction is validated in ten different cancer datasets. We use breast cancer and glioblastoma cohorts as examples to show how subtype specific drug sensitivity can be translated to individual tumour samples. The low dimensional latent representations extracted by MFmap explain known and novel subtype specific features and enable the analysis of cell-state transformations between different subtypes. From a methodological perspective, we report that MFmap is a semi-supervised method which simultaneously achieves good generative and predictive performance and thus opens opportunities in other areas of computational biology.

MFmap: A semi-supervised generative model matching cell lines to tumours and cancer subtypes

Translating in vitro results from experiments with cancer cell lines to clinical applications requires the selection of appropriate cell line models. Here we present MFmap (model fidelity map), a machine learning model to simultaneously predict the cancer subtype of a cell line and its similarity to an individual tumour sample. The MFmap is a semi-supervised generative model, which compresses high dimensional gene expression, copy number variation and mutation data into cancer subtype informed low dimensional latent representations. The accuracy (test set F 1 score > 90%) of the MFmap subtype prediction is validated in ten different cancer datasets. We use breast cancer and glioblastoma cohorts as examples to show how subtype specific drug sensitivity can be translated to individual tumour samples. The low dimensional latent representations extracted by MFmap explain known and novel subtype specific features and enable the analysis of cell-state transformations between different subtypes. From a methodological perspective, we report that MFmap is a semi-supervised method which simultaneously achieves good generative and predictive performance and thus opens opportunities in other areas of computational biology.

Stereotactic body radiotherapy with a single isocentre for multiple pulmonary metastases

A 45-year-old male developed a second set of pulmonary metastases 5 years after surgery for extraskeletal mucinous chondrosarcoma of the left shoulder. He already underwent a lobectomy and two segmentectomies for a first set of pulmonary metastases 2 years ago. The closely grouped three nodules within the left lower lung formed a planning target volume (PTV) for stereotactic body radiotherapy (SBRT) with a single isocentre, which was focused on the centre of the largest nodule (the simultaneous plan). Dose-volume histogram analysis confirmed that the plan was superior to an alternative plan, in which SBRT plans would have been produced for each individual tumour (the individual plan). The mean, maximum and minimum PTV doses were 54.0, 57.5 and 47.3 Gy, respectively, in the simultaneous plan, and 65.6, 87.2 and 52.3 Gy, respectively, in the individual plan. The homogeneity index, conformity index, and the maximum dose delivered to the surrounding healthy lung were 1.21, 0.71, and 37.7 Gy, respectively, in the simultaneous plan and 1.66, 4.44, and 46.2 Gy, respectively, in the individual plan. The patient developed Grade two pneumonitis, but remained healthy until 4 years after the SBRT. When multiple closely grouped metastases are treated using SBRT, the use of a single isocentre should be considered.

Imaging in Paediatric Oncology

All superficial and abdominal masses in children should be first evaluated with ultrasound including Doppler assessment for lesion vascularity. Many benign and all malignant masses will need further evaluation with CT or MRI. For non-CNS tumours, MRI is preferred over CT, if feasible, bearing in mind anaesthesia may be needed for children under seven years of age. MRI is preferable to CT because of a lack of ionizing radiation, better soft-tissue evaluation in general, improved depiction of possible bone marrow disease, and, where present, better demonstration of spinal cord compression. Most non-CNS tumours, with the exception of neuroblastoma, would merit a routine CT of the chest for disease staging. Characteristics of individual tumour types, and the emerging role of nuclear medicine in staging and response assessment, will be mentioned in the chapter.

Assessment of 5-methylcytosine and 5-hydroxymethylcytosine technologies in contrived ctDNA-like samples.

e15585 Background: Epigenomic changes to DNA are early driving events in tumorigenesis and are specific markers of the tumour and its microenvironment. Cell-free DNA that is released from the tumour microenvironment can potentially be used for the detection of tumour-specific epigenomic changes for cancer liquid biopsy testing. In this study, we assessed the robustness of 5-methylcytosine and 5-hydroxymethylcytosine technologies in contrived ctDNA-like samples from NSCLC. Additionally, we assessed the limit of detection of the tumour fraction within the tested dilution range looking at a subset of differential epigenomic marks. Methods: 4 NSCLC tumour and buffy coat DNA samples were commercially obtained and processed to generate contrived tumour DNA dilution mimicking circulating tumour DNA at varying tumour fractions. Samples were sent in triplicate to multiple epigenomic assay providers for analysis. The robustness i.e. reproducibility and limit-of-detection within the tested range was determined. Results: Technical replicates all passed method-specific quality controls and were highly reproducible. Cancer specific differentially-methylated or hydroxymethylated regions were determined for each individual tumour sample as a “sample-specific cancer signature”; when applied to the tumour dilution samples, these signatures were capable of distinguishing all tumour dilutions down to 0.05% from normal background DNA. Detection of differential methylation and hydroxymethylation regions was reproducible across technical replicates for all four NCSLC samples at > 0.05%. Generation of pooled methylation “signatures” derived from commercially available healthy plasma and cancer tissue datasets could still detect cancer DNA in these four NCSLC samples at > 0.1% tumour fraction. Conclusions: Our pilot study showed that all assessed technologies generated reproducible, robust data and could detect ctDNA at even clinically relevant levels for early stage cancer. One limitation of this study was the small number of “n” and the use of sonicated contrived sample that may induce artificial bias and reduce sensitivity highlighting a requirement for the development of reference material in the community. The ability to detect cancer-specific epigenomic changes from liquid biopsy with high sensitivity offers new promising avenues for early detection of cancer, which can increase the chances of successful treatment.

Learning mutational graphs of individual tumour evolution from single-cell and multi-region sequencing data

BackgroundA large number of algorithms is being developed to reconstruct evolutionary models of individual tumours from genome sequencing data. Most methods can analyze multiple samples collected either through bulk multi-region sequencing experiments or the sequencing of individual cancer cells. However, rarely the same method can support both data types.ResultsWe introduce TRaIT, a computational framework to infer mutational graphs that model the accumulation of multiple types of somatic alterations driving tumour evolution. Compared to other tools, TRaIT supports multi-region and single-cell sequencing data within the same statistical framework, and delivers expressive models that capture many complex evolutionary phenomena. TRaIT improves accuracy, robustness to data-specific errors and computational complexity compared to competing methods.ConclusionsWe show that the application of TRaIT to single-cell and multi-region cancer datasets can produce accurate and reliable models of single-tumour evolution, quantify the extent of intra-tumour heterogeneity and generate new testable experimental hypotheses.

Molecular Diagnostics in Breast Cancer Routine Practice

The portfolio of adjuvant systemic treatment of breast cancer nowadays contains novel anti-hormonal and chemotherapeutic drugs, immunotherapeutic approaches and small molecules that are only effective in a limited number of patients and are often associated with high costs and significant side effects. Therefore, a personalised approach based on individual tumour biomarkers is required to arrive at the optimal balance between effectiveness on the one hand, and costs and side effects on the other. The aim of this paper is to provide an overview of the molecular biomarkers and associated molecular tests that are currently relevant in pathology of invasive breast cancer.

Long lasting octreotide LAR therapy in large cell neuroendocrine carcinoma (LCNEC) of the lung

In this article we report a case of a patient with large cell neuroendocrine carcinoma (LCNEC) of the lung. Patients with LCNEC usually have poor prognosis and the benefit of adjuvant chemotherapy for these patients has not been fully established. This case suggests that octreotide LAR, a somatostatine analogue (SSA), can be useful in the treatment of neuroendocrine carcinoma also as maintenance therapy in association with chemotherapy. Further studies, regarding individual tumour biological behaviour and SSAs optimal dosage, could be useful to optimise treatment and to add new insights into the mechanisms of action and the role of SSAs in the therapy of NETs.