Sparse Regression in Cancer Genomics: Comparing Variable Selection and Predictions in Real World Data.

BACKGROUND: Evaluation of gene interaction models in cancer genomics is challenging, as the true distribution is uncertain. Previous analyses have benchmarked models using synthetic data or databases of experimentally verified interactions - approaches which are susceptible to misrepresentation and incompleteness, respectively. The objectives of this analysis are to (1) provide a real-world data-driven approach for comparing performance of genomic model inference algorithms, (2) compare the performance of LASSO, elastic net, best-subset selection, L 0 L 1 penalisation and L 0 L 2 penalisation in real genomic data and (3) compare algorithmic preselection according to performance in our benchmark datasets to algorithmic selection by internal cross-validation. METHODS: Five large ( n 4000 ) genomic datasets were extracted from Gene Expression Omnibus. 'Gold-standard' regression models were trained on subspaces of these datasets ( n 4000 , p = 500 ). Penalised regression models were trained on small samples from these subspaces ( n ∈ { 25 , 75 , 150 } , p = 500 ) and validated against the gold-standard models. Variable selection performance and out-of-sample prediction were assessed. Penalty 'preselection' according to test performance in the other 4 datasets was compared to selection internal cross-validation error minimisation. RESULTS: L 1 L 2 -penalisation achieved the highest cosine similarity between estimated coefficients and those of gold-standard models. L 0 L 2 -penalised models explained the greatest proportion of variance in test responses, though performance was unreliable in low signal:noise conditions. L 0 L 2 also attained the highest overall median variable selection F1 score. Penalty preselection significantly outperformed selection by internal cross-validation in each of 3 examined metrics. CONCLUSIONS: This analysis explores a novel approach for comparisons of model selection approaches in real genomic data from 5 cancers. Our benchmarking datasets have been made publicly available for use in future research. Our findings support the use of L 0 L 2 penalisation for structural selection and L 1 L 2 penalisation for coefficient recovery in genomic data. Evaluation of learning algorithms according to observed test performance in external genomic datasets yields valuable insights into actual test performance, providing a data-driven complement to internal cross-validation in genomic regression tasks.

Preclinical development and characterisation of 99mTc-NM-01 for SPECT/CT imaging of human PD-L1.

The level of expression of programmed cell death-1 (PD-1)/programmed death ligand-1 (PD-L1) is a predictive biomarker for cancer immunotherapy, however, its detection remains challenging due to tumour heterogeneity and the influence from the binding of therapeutic agents. We recently developed [99mTc]-NM-01 as a companion diagnostic imaging agent for non-invasive molecular imaging of PD-L1 by single-photon emission computed tomography (SPECT). The aim of the study was to evaluate the [99mTc] radiolabelling of GMP graded NM-01 and its pharmacology, pharmacokinetics and toxicology. NM-01 bound specifically to human PD-L1 (Kd=0.8 nM) and did not interfere with the binding of the anti-PD-L1 antibody atezolizumab. NM-01 can bind various PD-L1-positive cancer cell lines and only interact with PD-L1 expressed on the cell surface. In SPECT/CT imaging, high [99mTc]-NM-01 accumulation was observed in the HCC827 mouse xenografted tumour model (30-min: 1.50 ± 0.27 %ID/g; 90-min: 1.23 ± 0.18 %ID/g), demonstrated a predominantly renal elimination (high uptake in bladder and kidney), while activity in the blood pool and other major organs remained low. The tumour-to-muscle and tumour-to-blood ratios were comparable with/without atezolizumab (P<0.04) but were significantly lowered when co-injected with excess NM-01 (P=0.04 and P=0.01, respectively.) The blood clearance of [99mTc]-NM-01 is bi-phasic; consisting of an initial fast washout phase with half-life of 2.1 min and a slower clearance phase with half-life of 25.4 min. In an intravenous extended single-dose toxicity study, no treatment-related changes were observed and the maximum tolerated dose of [99mTc]-NM-01 was 2.58 mg/kg. [99mTc]-NM-01 has suitable properties as a potential candidate for SPECT/CT imaging of PD-L1 assessment in cancer patients.

Functional and Hybrid Imaging of Bone Metastases.

Bone metastases are common, cause significant morbidity, and impact on healthcare resources. Although radiography, computed tomography (CT), magnetic resonance imaging (MRI), and bone scintigraphy have frequently been used for staging the skeleton, these methods are insensitive and nonspecific for monitoring treatment response in a clinically relevant time frame. We summarize several recent reports on new functional and hybrid imaging methods including single photon emission CT/CT, positron emission tomography/CT, and whole-body MRI with diffusion-weighted imaging. These modalities generally show improvements in diagnostic accuracy for staging and response assessment over standard imaging methods, with the ability to quantify biological processes related to the bone microenvironment as well as tumor cells. As some of these methods are now being adopted into routine clinical practice and clinical trials, further evaluation with comparative studies is required to guide optimal and cost-effective clinical management of patients with skeletal metastases. © 2018 American Society for Bone and Mineral Research.

Magnetic Resonance Imaging (MRI) of Intratumoral Voxel Heterogeneity as a Potential Response Biomarker: Assessment in a HER2+ Esophageal Adenocarcinoma Xenograft Following Trastuzumab and/or Cisplatin Therapy.

We evaluated magnetic resonance imaging (MRI) voxel heterogeneity following trastuzumab and/or cisplatin in a HER2+ esophageal xenograft (OE19) as a potential response biomarker. OE19 xenografts treated with saline (controls), monotherapy, or combined cisplatin and trastuzumab underwent 9.4-T MRI. Tumor MRI parametric maps of T1 relaxation time (pre/post contrast), T2 relaxation time, T2* relaxation rate (R2*), and apparent diffusion coefficient obtained before (TIME0), after 24hours (TIME1), and after 2weeks of treatment (TIME2) were analyzed. Voxel histogram and fractal parameters (from the whole tumor, rim and center, and as a ratio of rim-to-center) were derived. Tumors were stained for immunohistochemical markers of hypoxia (CA-IX), angiogenesis (CD34), and proliferation (Ki-67). Combination therapy reduced xenograft growth rate (relative change, ∆ +0.58±0.43 versus controls, ∆ +4.1±1.0; P=0.008). More spatially homogeneous voxel distribution between the rim to center was noted after treatment for combination therapy versus controls, respectively, for contrast-enhanced T1 relaxation time (90th percentile: ratio 1.00 versus 0.88, P=0.009), T2 relaxation time (mean: 1.00 versus 0.92, P=0.006; median: 0.98 versus 0.91, P=0.006; 75th percentile: 1.02 versus 0.94, P=0.007), and R2* (10th percentile: 0.99 versus 1.26, P=0.003). We found that combination and trastuzumab monotherapy reduced MRI spatial heterogeneity and growth rate compared to the control or cisplatin groups, the former providing adjunctive tumor response information.