We at mBiosphere know you are busy, reader! With various gels, analyses, programs, and classes to run, not to mention reports, abstracts, and grants to write, we know there are many demands made on our readers’ time (plus, dinner to plan, laundry to fold, the dog to walk...wait! Don't go! You have time to read this, I swear!). The smart folks at mSystems also recognize their readers’ full schedules, and thus they’ve initiated a new video introduction system in which authors themselves give a short summation of their recently published reports. The videos feature the authors providing a clear, condensed explanation of why and how the research was performed.
The first video introduction is from Casey Green, a researcher who understands the value in brevity (he has a tl;dr summary at the top of his lab website). Green begins by giving background to a paper published by first author Jie Tan in his lab, in collaboration with Deborah Hogan:
In the video, Green explains the difficult problem of building systems-level tools that can sift through large amounts of data to separate background noise from significant findings. The software system described in the recently published mSystems article, analysis using denoising autoencoders of gene expression (ADAGE), was used to identify common genes altered in numerous Pseudomonas aeruginosa experiments.
ADAGE analyzed a wide number of published microarray and transcriptome sequencing results to find consistent patterns. Genes were associated with nodes, which indicate a biological characteristic (e.g., a particular growth condition or P. aeruginosa strain), with some genes more heavily weighted in a given node. Left, you see the genes associated with a particular node, with the heavily weighted genes of two different nodes comprising the two inner circles. The ADAGE P. aeruginosa network was then taught to ignore noise by adding in artificial background and instructing the network to remove this noise, increasing its ability to identify true associations.
After establishing the ADAGE relationships, the system was tested. First the researchers looked for association between nodes and genes located within the same operon. Operons are bacterial gene sets coregulated by a similar promoter, and were therefore expected to be on or off under the same node conditions. Testing this idea using the ADAGE networked showed that it panned out: if one operonic gene was highly weighted in association with a particular node, other genes within that operon were more likely to have a similar association. The same was true for spatially proximal genes – genes nearby each other on the chromosome, regardless of operon status.
ADAGE differentiation between P. aeruginosa strains was examined by looking more closely at the differences in gene associations between one node and two commonly studied strains, PAO1 and PA14. This node contained several strain-specific genes, including genes involved in lipopolysaccharide biosynthesis and a bacteriophage found only in certain P. aeruginosa strains. By analyzing published experiments of multiple strains, they researchers observed that this node was differentially active when comparing these two strains.
With proof-of-principle in hand, the researchers speculate that ADAGE models can be a source of biological discovery through application to already-generated data. The ability of the network to filter noise allows comparison of gene expression data generated in different labs, different strains, and different growth conditions. This type of systemic tool will only become more valuable as next-generation sequencing leads to stockpiles of gene expression datasets, which in turn can be used to refine the ADAGE network.
This type of analysis adds value to the many genome-wide data sets looking at transcriptional differences between defined strains or during distinct environmental conditions. ADAGE will not only allow analysis of many datasets concurrently, it also saves researchers the valuable time it would take to generate the datasets in the first place. And as established earlier, time-saving tools are something everyone can appreciate!
-- Julie Wolf