Deep learning for computational biology

Technological advances in genomics and imaging have led to an explosion of molecular and cellular profiling data from large numbers of samples. This rapid increase in biological data dimension and acquisition rate is challenging conventional analysis strategies. Modern machine learning methods, such as deep learning, promise to leverage very large data sets for finding hidden structure within them, and for making accurate predictions. In this review, we discuss applications of this new breed of analysis approaches in regulatory genomics and cellular imaging. We provide background of what deep learning is, and the settings in which it can be successfully applied to derive biological insights. In addition to presenting specific applications and providing tips for practical use, we also highlight possible pitfalls and limitations to guide computational biologists when and how to make the most use of this new technology.     

Water masses in the Bransfield Strait and adjacent seas,austral summer 2013

The Bransfield Strait is a semi-enclosed sea located in the northern part of the West Antarctic Peninsula region, which is subject to strong climatic changes. The bathymetry is complex and comprises three basins that are separated from each other by shallow sills. Oceanographic measurements of the Bransfield Strait region are available since the first half of the twentieth century. In this study, hydrographic data from the ANT-XXIX/3 expedition of RV Polarstern in 2013 are presented to describe the actual physical state of the art, particularly for biological work done during that cruise. The general hydrographic situation of the Bransfield Strait in 2013 is found to be similar to observations from the early twentieth century. The Bransfield Strait’s water masses are modified versions of the water masses from the adjacent seas. The different water masses within the Bransfield Strait are separated by two fronts, the so-called Bransfield and Peninsula Front. While the Bransfield Front is most pronounced in the central and southwestern Bransfield Strait, the Peninsula Front can be identified from the northeastern to the central part of the study domain. Based on an analysis of water mass properties around the Antarctic Peninsula and close to the Antarctic Sound, a notable inflow of Shelf Water from the Weddell Sea through the Antarctic Sound appears unlikely.     

Crystal structure and functional analysis of MiD49, a receptor for the mitochondrial fission protein Drp1

Mitochondrial fission requires recruitment of dynamin‐related protein 1 (Drp1) to the mitochondrial surface, where assembly leads to activation of its GTP‐dependent scission function. MiD49 and MiD51 are two receptors on the mitochondrial outer membrane that can recruit Drp1 to facilitate mitochondrial fission. Structural studies indicated that MiD51 has a variant nucleotidyl transferase fold that binds an ADP co‐factor essential for activation of Drp1 function. MiD49 shares sequence homology with MiD51 and regulates Drp1 function. However, it is unknown if MiD49 binds an analogous co‐factor. Because MiD49 does not readily crystallize, we used structural predictions and biochemical screening to identify a surface entropy reduction mutant that facilitated crystallization. Using molecular replacement, we determined the atomic structure of MiD49 to 2.4 Å. Like MiD51, MiD49 contains a nucleotidyl transferase domain; however, the electron density provides no evidence for a small‐molecule ligand. Structural changes in the putative nucleotide‐binding pocket make MiD49 incompatible with an extended ligand like ADP, and critical nucleotide‐binding residues found in MiD51 are not conserved. MiD49 contains a surface loop that physically interacts with Drp1 and is necessary for Drp1 recruitment to the mitochondrial surface. Our results suggest a structural basis for the differential regulation of MiD51‐ versus MiD49‐mediated fission.     

Carbon uptake in Eurasian boreal forests dominates the high-latitude net ecosystem carbon budget

Arctic-boreal landscapes are experiencing profound warming, along with changes in ecosystem moisture status and disturbance from fire. This region is of global importance in terms of carbon feedbacks to climate, yet the sign (sink or source) and magnitude of the Arctic-boreal carbon budget within recent years remains highly uncertain. Here, we provide new estimates of recent (2003–2015) vegetation gross primary productivity (GPP), ecosystem respiration (R_eco), net ecosystem CO₂ exchange (NEE; R_eco − GPP), and terrestrial methane (CH₄) emissions for the Arctic-boreal zone using a satellite data-driven process-model for northern ecosystems (TCFM-Arctic), calibrated and evaluated using measurements from >60 tower eddy covariance (EC) sites. We used TCFM-Arctic to obtain daily 1-km² flux estimates and annual carbon budgets for the pan-Arctic-boreal region. Across the domain, the model indicated an overall average NEE sink of −850 Tg CO₂-C year⁻¹. Eurasian boreal zones, especially those in Siberia, contributed to a majority of the net sink. In contrast, the tundra biome was relatively carbon neutral (ranging from small sink to source). Regional CH₄ emissions from tundra and boreal wetlands (not accounting for aquatic CH₄) were estimated at 35 Tg CH₄-C year⁻¹. Accounting for additional emissions from open water aquatic bodies and from fire, using available estimates from the literature, reduced the total regional NEE sink by 21% and shifted many far northern tundra landscapes, and some boreal forests, to a net carbon source. This assessment, based on in situ observations and models, improves our understanding of the high-latitude carbon status and also indicates a continued need for integrated site-to-regional assessments to monitor the vulnerability of these ecosystems to climate change.     

SH3P7 Is a Cytoskeleton Adapter Protein and Is Coupled to Signal Transduction from Lymphocyte Antigen Receptors

Lymphocytes respond to antigen receptor engagement with tyrosine phosphorylation of many cellular proteins, some of which have been identified and functionally characterized. Here we describe SH3P7, a novel substrate protein for Src and Syk family kinases. SH3P7 migrates in sodium dodecyl sulfate-polyacrylamide gel electrophoresis as a 55-kDa protein that is preferentially expressed in brain, thymus, and spleen. It contains multiple amino acid sequence motifs, including two consensus tyrosine phosphorylation sites of the YXXP type and one SH3 domain. A region of sequence similarity, which we named SCAD, was found in SH3P7 and three actin-binding proteins. The SCAD region may represent a new type of protein-protein interaction domain that mediates binding to actin. Consistent with this possibility, SH3P7 colocalizes with actin filaments of the cytoskeleton. Altogether, our data implicate SH3P7 as an adapter protein which links antigen receptor signaling to components of the cytoskeleton.     

Analysis of plasmid deoxyribonucleic acid in a cariogenic strain of Streptococcus faecalis: an approach to identifying genetic determinants on cryptic plasmids.

Streptococcus faecalis strains ND539 and OG1 have been previously shown to be cariogenic in gnotobiotic animals. Deoxyribonucleic acid analyses have revealed the presence of a single 26-megadalton plasmid designated pAM539 in the former strain, whereas the latter strain was found to be plasmid-free. By gene transfer experiments, it was possible to construct isogenic pairs of strains that differed only with regard to the presence or absence of pAM539. Comparative studies of isogenic pairs showed that the presence of pAM539 conferred bacterial sensitivity to a bacteriocin produced by S. faecalis strain 5952.     

Analytical and Clinical Validation of a Digital Sequencing Panel for Quantitative,Highly Accurate Evaluation of Cell-Free Circulating Tumor DNA

Next-generation sequencing of cell-free circulating solid tumor DNA addresses two challenges in contemporary cancer care. First this method of massively parallel and deep sequencing enables assessment of a comprehensive panel of genomic targets from a single sample, and second, it obviates the need for repeat invasive tissue biopsies. Digital Sequencing^TM is a novel method for high-quality sequencing of circulating tumor DNA simultaneously across a comprehensive panel of over 50 cancer-related genes with a simple blood test. Here we report the analytic and clinical validation of the gene panel. Analytic sensitivity down to 0.1% mutant allele fraction is demonstrated via serial dilution studies of known samples. Near-perfect analytic specificity (> 99.9999%) enables complete coverage of many genes without the false positives typically seen with traditional sequencing assays at mutant allele frequencies or fractions below 5%. We compared digital sequencing of plasma-derived cell-free DNA to tissue-based sequencing on 165 consecutive matched samples from five outside centers in patients with stage III-IV solid tumor cancers. Clinical sensitivity of plasma-derived NGS was 85.0%, comparable to 80.7% sensitivity for tissue. The assay success rate on 1,000 consecutive samples in clinical practice was 99.8%. Digital sequencing of plasma-derived DNA is indicated in advanced cancer patients to prevent repeated invasive biopsies when the initial biopsy is inadequate, unobtainable for genomic testing, or uninformative, or when the patient’s cancer has progressed despite treatment. Its clinical utility is derived from reduction in the costs, complications and delays associated with invasive tissue biopsies for genomic testing.     

Self-eating to grow and kill: autophagy in filamentous ascomycetes

Autophagy is a tightly controlled degradation process in which eukaryotic cells digest their own cytoplasm containing protein complexes and organelles in the vacuole or lysosome. Two types of autophagy have been described: macroautophagy and microautophagy. Both types can be further divided into nonselective and selective processes. Molecular analysis of autophagy over the last two decades has mostly used the unicellular ascomycetes Saccharomyces cerevisiae and Pichia pastoris. Genetic analysis in these yeasts has identified 36 autophagy-related (atg) genes; many are conserved in all eukaryotes, including filamentous ascomycetes. However, the autophagic machinery also evolved significant differences in fungi, as a consequence of adaptation to diverse fungal lifestyles. Intensive studies on autophagy in the last few years have shown that autophagy in filamentous fungi is not only involved in nutrient homeostasis but in other cellular processes such as cell differentiation, pathogenicity and secondary metabolite production. This mini-review focuses on the specific roles of autophagy in filamentous fungi.