ABSTRACT. Double minute chromosomes are highly amplified oncogenic, acentric, extrachromosomal DNA that are frequently observed in the cells of numerous cancer types. Algorithmic discovery of double minutes (DM) can potentially improve bench-derived therapies for cancer treatment. A hindrance to this task is that DMs evolve, yielding circular chromatin that shares segments from progenitor double minutes. This creates multiple double minutes in the tumor genome that are distinct, but that share loci for overlapping amplicon coordinates. Existing DM discovery algorithms (that use only whole genome sequencing data) can potentially misclassify DMs that share overlapping coordinates. In this study, we describe a method called HolistIC that predicts double minutes in tumor genomes by integrating whole genome shotgun sequencing (WGS) and Hi-C sequencing data. This resolves ambiguity in double minute prediction that exists when using WGS data alone, a limitation of existing methods for this problem. We implemented and tested our algorithm on the tandem Hi-C and WGS datasets of two cancer datasets and a simulated dataset. Our results show that HolistIC can distinguish between double minutes that share amplicon coordinates, an advance over current methods for this problem.

ABSTRACT. There is a growing interest in using computational methods to screen for peptides that can be used as targets for cancer immunotherapy, and structure-based methods are particularly interesting for personalized predictions. However, computationally docking peptides to protein receptors remains an open challenge. Part of the issue lies in accurately scoring the binding modes of a protein-peptide complex produced by a molecular docking tool. In this study, we evaluate several popular scoring functions in their ability to accurately rank the best protein-peptide complex conformations, based on the RMSD to the original crystal structure. Our evaluation of scoring shows that existing scoring functions, though known to be adept at scoring drug-like ligands, are limited when it comes to peptides. Therefore, accurately scoring conformations from peptide docking should currently be regarded as the biggest unmet challenge in molecular docking and as an important goal for computational oncology.

ABSTRACT. Gastric cancer is the sixth most common worldwide malignancy and the third leading cancer cause of death. Early diagnosis and effective after-surgical monitoring can significantly improve survival rates. Previous studies have revealed several serum biomarkers that are elevated in gastric cancer patients, including CEA, CA19-9, and CA72-4. However, sensitivity of these biomarkers is below 30%. Identification of more sensitive gastric cancer markers is critical for individualized gastric cancer therapy. Here we developed an approach for single-cell transcriptomic data analysis that identifies secretory proteins, which would be measurable in the blood, that are abundantly expressed in cancer cells. Using gastric cancer scRNA-seq data, we identified 19 secretory proteins that could be used in a gastric cancer diagnostic panel. Moreover, three of these markers KLK7, CFD and F12 that are well-known to be involved in tumor microenvironment formation, were not previously associated with gastric cancer. If verified, our data suggests a novel assay for the presence/absence of occult metastases.

ABSTRACT. Cancer is complicated and complex, arising in multiple cell types and tissue of origin, initiating and progressing differently in different individuals or under different conditions, with effects crossing multiple biological scales.  It is also adaptive, evolving at both the molecular and cellular scales during treatment.  Because of this, cancer researchers are often early adopters of new technologies, methods and approaches and they routinely adopt tools and methods originally developed in other, very different fields for use in modeling, understanding, and combating cancer.  Systems biology, with its combination of experimental biology and mathematical modeling, plays an increasingly important role in cancer research.  Advances in data science, high performance computing, and artificial intelligence have led to increasingly creative adoption of these tools and approaches in cancer research.  But high-dimensional data, complicated problems, and collaborative problem solving also exist in areas such as entertainment video games and virtual and augmented reality.  And cancer data and its contextual information can come from a variety of sources including patients and the public through citizen science and crowdsourcing.  This talk will explore this broader collaborative and creative space and high-light both on-going and new programs supported by the National Cancer Institute to create opportunities for cross-field interactions and to accelerate new collaborations and new inter-disciplinary approaches to cancer research.