Comprehensible Predictive Modeling Using Regularized Logistic Regression and Comorbidity Based Features

Different studies have demonstrated the importance of comorbidities to better understand the origin and evolution of medical complications. This study focuses on improvement of the predictive model interpretability based on simple logical features representing comorbidities. We use group lasso based feature interaction discovery followed by a post-processing step, where simple logic terms are added. In the final step, we reduce the feature set by applying lasso logistic regression to obtain a compact set of non-zero coefficients that represent a more comprehensible predictive model. The effectiveness of the proposed approach was demonstrated on a pediatric hospital discharge dataset that was used to build a readmission risk estimation model. The evaluation of the proposed method demonstrates a reduction of the initial set of features in a regression model by 72%, with a slight improvement in the Area Under the ROC Curve metric from 0.763 (95% CI: 0.755–0.771) to 0.769 (95% CI: 0.761–0.777). Additionally, our results show improvement in comprehensibility of the final predictive model using simple comorbidity based terms for logistic regression.     

High-Resolution Physical Map of the X-linked Retinoschisis Interval in Xp22

X-linked retinoschisis (RS) is the leading cause of macular degeneration in young males and has been mapped to Xp22 between DXS418 and DXS999. To facilitate identification of the RS gene, we have constructed a yeast artificial chromosome (YAC) contig across this region comprising 28 YACs and 32 sequence-tagged sites including seven novel end clone markers. To establish the definitive marker order, a PAC contig containing 50 clones was also constructed, and all clones were fingerprinted. The marker order is: Xpter–DXS1317–(AFM205yd12–DXS7175–DXS7992)–60N8T7–DXS1195–DXS7993–DXS7174–60N8-SP6–DXS418–DXS7994–DXS7995–DXS7996–HYAT2–25HA10R–HYAT1–DXS7997–DXS7998–DXS257–434E8R–3542R–DXS6762–DXS7999–DXS6763–434E8L–DXS8000–DXS6760–DXS7176–DXS8001–DXS999–3176R–PHKA2–Xcen. A long-range restriction map was constructed, and the RS region is estimated to be 1300 kb, containing three putative CpG islands. An unstable region was identified between DXS6763 and 434E8L. These data will facilitate positional cloning of RS and other disease genes in Xp22.     

Local and regional assessments of the impacts of plant invaders on vegetation structure and soil properties of Mediterranean islands

Aims Although biological invasions occur throughout the world, and some invaders are widespread in many habitats, few studies on the ecological impact of invaders have examined multiple sites. We tested how the impact of three widespread plant invaders changed depending on the identity of the species and the invaded island. We also tested whether relative species loss was lower in species‐rich communities than in species‐poor ones. Location We conducted floristic surveys and soil analyses in eight Mediterranean Basin islands: Crete and Lesbos (Greece), Sardinia (Italy), Corsica, Bagaud and Porquerolles (France), and Mallorca and Menorca (Spain). Methods We compared native species richness and diversity, proportion of life forms, soil percentage nitrogen, percentage organic carbon, C/N, and soil pH in nearby paired plots of 2 × 2 m: one control and one invaded by either the deciduous tree Ailanthus altissima, the succulent subshrubs Carpobrotus spp. or the annual geophyte Oxalis pes‐caprae, across eight Mediterranean Basin islands. Results On average, the presence of invaders reduced species diversity, Carpobrotus spp. exhibiting the largest impact and Oxalis the least. However, the relative impact was island‐dependent, and was positively but weakly associated with the species richness of the recipient community. Therophytes were the life form that experienced the largest decrease across islands. The effects of invasion on soil properties were very variable. Total N changed (increased) only in plots invaded by Ailanthus, significantly decreasing the C/N ratio. The presence of this tree increased soil pH, whereas the opposite was found in plots invaded by the other two species. Organic C increased in plots invaded by Ailanthus and Carpobrotus species. Main conclusions By conducting an analysis at multiple sites, we found that the three plant invaders had an impact on plant community structure not entirely concordant with changes in soil properties. The impacts depended on the identity of the species and of the invaded island, suggesting that impact of invaders is context‐specific. The impact in terms of species loss was not lower in species‐rich than in species‐poor communities.     

Specific Inter-residue Interactions as Determinants of Human Monoacylglycerol Lipase Catalytic Competency: A ROLE FOR GLOBAL CONFORMATIONAL CHANGES*

The serine hydrolase monoacylglycerol lipase (MGL) functions as the main metabolizing enzyme of 2-arachidonoyl glycerol, an endocannabinoid signaling lipid whose elevation through genetic or pharmacological MGL ablation exerts therapeutic effects in various preclinical disease models. To inform structure-based MGL inhibitor design, we report the direct NMR detection of a reversible equilibrium between active and inactive states of human MGL (hMGL) that is slow on the NMR time scale and can be modulated in a controlled manner by pH, temperature, and select point mutations. Kinetic measurements revealed that hMGL substrate turnover is rate-limited across this equilibrium. We identify a network of aromatic interactions and hydrogen bonds that regulates hMGL active-inactive state interconversion. The data highlight specific inter-residue interactions within hMGL modulating the enzymes function and implicate transitions between active (open) and inactive (closed) states of the hMGL lid domain in controlling substrate access to the enzymes active site.     

Correlation of bistranded clustered abasic DNA lesion processing with structural and dynamic DNA helix distortion

Clustered apurinic/apyrimidinic (AP; abasic) DNA lesions produced by ionizing radiation are by far more cytotoxic than isolated AP lesion entities. The structure and dynamics of a series of seven 23-bp oligonucleotides featuring simple bistranded clustered damage sites, comprising of two AP sites, zero, one, three or five bases 3′ or 5′ apart from each other, were investigated through 400 ns explicit solvent molecular dynamics simulations. They provide representative structures of synthetically engineered multiply damage sites-containing oligonucleotides whose repair was investigated experimentally (Nucl. Acids Res. 2004, 32:5609-5620; Nucl. Acids Res. 2002, 30: 2800–2808). The inspection of extrahelical positioning of the AP sites, bulge and non Watson–Crick hydrogen bonding corroborates the experimental measurements of repair efficiencies by bacterial or human AP endonucleases Nfo and APE1, respectively. This study provides unprecedented knowledge into the structure and dynamics of clustered abasic DNA lesions, notably rationalizing the non-symmetry with respect to 3′ to 5′ position. In addition, it provides strong mechanistic insights and basis for future studies on the effects of clustered DNA damage on the recognition and processing of these lesions by bacterial or human DNA repair enzymes specialized in the processing of such lesions.     

Nodule-inducing activity of synthetic Sinorhizobium meliloti nodulation factors and related lipo-chitooligosaccharides on alfalfa. Importance of the acyl chain structure.

Sinorhizobium meliloti nodulation factors (NFs) elicit a number of symbiotic responses in alfalfa (Medicago sativa) roots. Using a semiquantitative nodulation assay, we have shown that chemically synthesized NFs trigger nodule formation in the same range of concentrations (down to 10(-10) M) as natural NFs. The absence of O-sulfate or O-acetate substitutions resulted in a decrease in morphogenic activity of more than 100-fold and approximately 10-fold, respectively. To address the question of the influence of the structure of the N-acyl chain, we synthesized a series of sulfated tetrameric lipo-chitooligosaccharides (LCOs) having fatty acids of different lengths and with unsaturations either conjugated to the carbonyl group (2E) or located in the middle of the chain (9Z). A nonacylated, sulfated chitin tetramer was unable to elicit nodule formation. Acylation with short (C8) chains rendered the LCO active at 10(-7) M. The optimal chain length was C16, with the C16-LCO being more than 10-fold more active than the C12- and C18-LCOs. Unsaturations were important, and the diunsaturated 2E,9Z LCO was more active than the monounsaturated LCOs. We discuss different hypotheses for the role of the acyl chain in NF perception.     

Total synthesis of epothilone E and related side-chain modified analogues via a Stille coupling based strategy.

A Stille coupling strategy has been utilized to complete a total synthesis of epothilone E from vinyl iodide 7 and thiazole-stannane 8h. The central core fragment 7 and its trans-isomer 11 were prepared from triene 15 using ring-closing metathesis (RCM), and were subsequently coupled to a variety of alternative stannanes to provide a library of epothilone analogues 18a-o and 19a-o. The Stille coupling approach was then used to prepare epothilone B analogues from the key macrolactone intermediate 24 which was itself synthesized by a macrolactonization based strategy.     

Human leptin induces angiogenesis in vivo

Leptin is an adipocyte-produced peptide, which plays a crucial role in the regulation of body weight. There is also evidence that leptin stimulates endothelial cell proliferation and the formation of capillary-like tubes in vitro. The disc angiogenesis system was used to measure the angiogenic effect of leptin in vivo. Discs containing 25, 50, 100 and 250 ng/ml of leptin were implanted subcutaneously in Wistar rats, removed after a growth period of 7 and 14 days, and compared with spontaneous growth controls and with positive controls containing equivalent doses of vascular endothelial growth factor (VEGF). Discs were examined morphologically for stroma and vessel development and by immunohistochemistry for quantitative evaluation of angiogenesis. The specificity of the angiogenic effect of leptin was tested by blocking leptin with a polyclonal anti-leptin antibody. Leptin induced a significant level of angiogenesis in a dose-dependent manner both at 7 and 14 days, with a peak at the dose of 100 ng/ml. The angiogenic activity of leptin was completely abolished by the anti-leptin neutralizing antibody. VEGF also induced significant dose-dependent angiogenesis at the same time points with a peak observed at a concentration of 100 ng/ml. The angiogenic response to leptin was significantly higher at 7 days compared with VEGF but not at the 14-day time point. In conclusion, leptin has a specific angiogenic effect in vivo, which is dose- and time-dependent in a concentration range of 25–250 ng/ml. This effect is equivalent to the angiogenic effect of VEGF but is evident earlier compared with VEGF.     

Approaches to dimensionality reduction in proteomic biomarker studies

Mass-spectra based proteomic profiles have received widespreadattention as potential tools for biomarker discovery and earlydisease diagnosis. A major data-analytical problem involvedis the extremely high dimensionality (i.e. number of featuresor variables) of proteomic data, in particular when the samplesize is small. This article reviews dimensionality reductionmethods that have been used in proteomic biomarker studies.It then focuses on the problem of selecting the most appropriatemethod for a specific task or dataset, and proposes method combinationas a potential alternative to single-method selection. Finally,it points out the potential of novel dimension reduction techniques,in particular those that incorporate domain knowledge throughthe use of informative priors or causal inference.