From space to biomedicine: Enabling biomarker data science in the cloud

NASA’s Jet Propulsion Laboratory (JPL) is advancing research capabilities for data science with two of the National Cancer Institute’s major research programs, the Early Detection Research Network (EDRN) and the Molecular and Cellular Characterization of Screen-Detected Lesions (MCL), by enabling data-driven discovery for cancer biomarker research. The research team pioneered a national data science ecosystem for cancer biomarker research to capture, process, manage, share, and analyze data across multiple research centers. By collaborating on software and data-driven methods developed for space and earth science research, the biomarker research community is heavily leveraging similar capabilities to support the data and computational demands to analyze research data. This includes linking diverse data from clinical phenotypes to imaging to genomics. The data science infrastructure captures and links data from over 1600 annotations of cancer biomarkers to terabytes of analysis results on the cloud in a biomarker data commons known as “LabCAS”. As the data increases in size, it is critical that automated approaches be developed to “plug” laboratories and instruments into a data science infrastructure to systematically capture and analyze data directly. This includes the application of artificial intelligence and machine learning to automate annotation and scale science analysis.

Forensic palaeontology: The Archaeoraptor forgery

The Archaeoraptor fossil was announced as a 'missing link' and purported to be possibly the best evidence since Archaeopteryx that birds did, in fact, evolve from certain types of carnivorous dinosaur. It reportedly came from Early Cretaceous beds of China that have produced other spectacular fossils transitional between birds and extinct non-avian dinosaurs. But Archaeoraptor was revealed to be a forgery in which bones of a primitive bird and a non-flying dromaeosaurid dinosaur had been combined. Here we use high-resolution X-ray computed tomography (CT) to determine the nature and extent of the forgery, as well as how it was built, by imaging the fracture pattern and distribution of materials through the entire specimen.

A genetic selection for isolating cDNAs encoding secreted proteins

We describe a simple, rapid technique for simultaneously isolating large numbers of cDNAs encoding secreted proteins. The technique makes use of a facile genetic selection performed in a strain of Saccharomyces cerevisiae deleted for its endogenous invertase gene. A cDNA cloning vector which carries a modified invertase gene lacking its leader sequence is used in conjunction with this strain. Heterologous secreted genes fused appropriately upstream of this defective invertase provide the necessary signals to restore secretion, allowing the yeast to grow on sugars such as sucrose or raffinose. This microbial growth selection facilitates scanning cDNA libraries containing millions of clones, enabling the wholesale identification of novel secreted proteins without the need for specific bioassays. The technique is similar to one previously described (Klein et al. (1996) Proc. Natl. Acad. Sci. USA 93, 7108-7113). We describe results using a cDNA library derived from activated human peripheral blood mononuclear cells (PBMC). Genes identified from this library encoded signal sequences of proteins of diverse structure, function, and cellular location such as cytokines, type 1 and type 2 transmembrane proteins, and proteins found in intracellular organelles. In addition, a number of novel secreted proteins were identified, including a chemokine and a novel G-protein-coupled receptor. Since signal sequences possess features conserved throughout evolution, the procedure can be used to isolate genes encoding secreted proteins from both eukaryotes and prokaryotes.

Transgenic remodeling of the contractile apparatus in the mammalian heart

The structure-function relationships of the sarcomeric proteins in the mammalian cardiac compartment remain ill-defined because of the lack of a suitable model in which they can be readily manipulated or exchanged in vivo. To establish the validity of the transgenic paradigm for remodeling the mammalian heart, the murine alpha -cardiac myosin heavy chain gene promoter was used to express a ventricular myosin light chain-2 transgene (MLC2v) in both the atria and ventricles of the adult animal. Expression resulted in high levels of the transgene's transcript in both compartments. In the ventricle, the transgene was expressed against the background expression of the normal isoform. In the atrium, the transgene's expression would be ectopic, in that normally, MLC2v expression is restricted to the ventricle. Ectopic expression of the transgene in the atria resulted in a complete replacement of the atrial myosin light chain-2 protein isoform, although the endogenous isoform's steady state transcript levels were unchanged. In contrast, ventricular expression of the transgene had no effect at the protein level, despite an eightfold increase in MLC2v transcript levels. The data show that sarcomeric protein stoichiometry is maintained rigorously via posttransciptional regulation and that protein replacement can be achieved through a single transgenic manipulation.

Postaxial polydactyly in forelimbs of CRABP-II mutant mice

The cytoplasmic retinoic acid (RA)-binding protein CRABP-II is expressed widely throughout early morphogenesis in mouse embryo, but its expression becomes more restricted as organogenesis progresses. CRABP-II expression remains strong in the developing limb bud suggesting a role for this protein in limb patterning. Here, we show that the CRABP-II promoter can direct expression of a lacZ transgene in a specific posterior domain during limb bud development. In order to investigate in more detail the role played by CRABP-II in RA signal transduction, we have also generated mice homozygous for a null mutation of this gene. CRABPII−/− mice are viable and fertile but show a developmental defect of the forelimb, specifically an additional, postaxial digit. This digit is generally, but not exclusively, limited to a single forepaw of an individual animal. The penetrance of the phenotype varies according to the genetic background, occurring most frequently on the inbred 129Sv background (50%), less frequently on the C57Bl/6 background (30%) and rarely on the outbred CD1 background (10%). This developmental abnormality implies a role for CRABP-II in normal patterning of the limb.

Transgenic indicator mice for studying activated retinoic acid receptors during development

Retinoic acid (RA) receptors (RARs) are ligand-inducible transcription factors that bind to specific DNA sequences associated with the regulatory regions of RA-regulatable genes. Since RA has been implicated as an important factor both in normal development and in teratological studies, one would like to have a model system that detects the presence of activated receptors during development. We have constructed a recombinant reporter gene that has three copies of the RA response element (RARE) from the RAR beta-2 promoter 5' to the herpes simplex virus thymidine kinase promoter; this regulatory region is coupled to the bacterial beta-galactosidase reporter gene. This construct was RA inducible in transient transfection assays in F9 embryonal carcinoma cells. Transgenic embryos with this reporter gene construct exhibited restricted and reproducible patterns of beta-galactosidase activity during embryogenesis, beginning between gestational ages day 7.5 and 8.5. At day 8.5, beta-galactosidase activity was detected in the closed neurotube and somites. Day 8.5 embryos, from pregnant females fed RA 14 hr earlier, exhibited a greater intensity and distribution of beta-galactosidase activity. Similarly, at later stages of gestation, maternal RA exposure resulted in enhanced embryonic beta-galactosidase expression. This type of transgenic indicator mouse should be useful in detailing the role of activated RARs during embryonic development.

Some requirements and suggestions for a methodology to develop knowledge based systems

This paper describes an approach to the creation of a methodology for the development of knowledge based systems. It specifies some requirements and suggests how these requirements might be met. General requirements can be satisfied using a systems approach. More specific ones can be met by viewing an organization as a network of consultations for coordinating expertise. The nature of consultations is described and the form of a possible cognitive model using a blackboard architecture is outlined. The value of the approach is illustrated in terms of certain knowledge elicitation methods.