Convolutional neural networks improve species distribution modelling by capturing the spatial structure of the environment

Convolutional Neural Networks (CNNs) are statistical models suited for learning complex visual patterns. In the context of Species Distribution Models (SDM) and in line with predictions of landscape ecology and island biogeography, CNN could grasp how local landscape structure affects prediction of species occurrence in SDMs. The prediction can thus reflect the signatures of entangled ecological processes. Although previous machine-learning based SDMs can learn complex influences of environmental predictors, they cannot acknowledge the influence of environmental structure in local landscapes (hence denoted “punctual models”). In this study, we applied CNNs to a large dataset of plant occurrences in France (GBIF), on a large taxonomical scale, to predict ranked relative probability of species (by joint learning) to any geographical position. We examined the way local environmental landscapes improve prediction by performing alternative CNN models deprived of information on landscape heterogeneity and structure (“ablation experiments”). We found that the landscape structure around location crucially contributed to improve predictive performance of CNN-SDMs. CNN models can classify the predicted distributions of many species, as other joint modelling approaches, but they further prove efficient in identifying the influence of local environmental landscapes. CNN can then represent signatures of spatially structured environmental drivers. The prediction gain is noticeable for rare species, which open promising perspectives for biodiversity monitoring and conservation strategies. Therefore, the approach is of both theoretical and practical interest. We discuss the way to test hypotheses on the patterns learnt by CNN, which should be essential for further interpretation of the ecological processes at play.     

Effect of Sequence and Stereochemistry Reversal on p53 Peptide Mimicry

Peptidomimetics effective in modulating protein-protein interactions and resistant to proteolysis have potential in therapeutic applications. An appealing yet underperforming peptidomimetic strategy is to employ D-amino acids and reversed sequences to mimic a lead peptide conformation, either separately or as the combined retro-inverso peptide. In this work, we examine the conformations of inverse, reverse and retro-inverso peptides of p53(15–29) using implicit solvent molecular dynamics simulation and circular dichroism spectroscopy. In order to obtain converged ensembles for the peptides, we find enhanced sampling is required via the replica exchange molecular dynamics method. From these replica exchange simulations, the D-peptide analogues of p53(15–29) result in a predominantly left-handed helical conformation. When the parent sequence is reversed sequence as either the L-peptide and D-peptide, these peptides display a greater helical propensity, feature reflected by NMR and CD studies in TFE/water solvent. The simulations also indicate that, while approximately similar orientations of the side-chains are possible by the peptide analogues, their ability to mimic the parent peptide is severely compromised by backbone orientation (for D-amino acids) and side-chain orientation (for reversed sequences). A retro-inverso peptide is disadvantaged as a mimic in both aspects, and further chemical modification is required to enable this concept to be used fruitfully in peptidomimetic design. The replica exchange molecular simulation approach adopted here, with its ability to provide detailed conformational insights into modified peptides, has potential as a tool to guide structure-based design of new improved peptidomimetics.     

Basal Ganglia Neuronal Activity during Scanning Eye Movements in Parkinson’s Disease

The oculomotor role of the basal ganglia has been supported by extensive evidence, although their role in scanning eye movements is poorly understood. Nineteen Parkinsońs disease patients, which underwent implantation of deep brain stimulation electrodes, were investigated with simultaneous intraoperative microelectrode recordings and single channel electrooculography in a scanning eye movement task by viewing a series of colored pictures selected from the International Affective Picture System. Four patients additionally underwent a visually guided saccade task. Microelectrode recordings were analyzed selectively from the subthalamic nucleus, substantia nigra pars reticulata and from the globus pallidus by the WaveClus program which allowed for detection and sorting of individual neurons. The relationship between neuronal firing rate and eye movements was studied by crosscorrelation analysis. Out of 183 neurons that were detected, 130 were found in the subthalamic nucleus, 30 in the substantia nigra and 23 in the globus pallidus. Twenty percent of the neurons in each of these structures showed eye movement-related activity. Neurons related to scanning eye movements were mostly unrelated to the visually guided saccades. We conclude that a relatively large number of basal ganglia neurons are involved in eye motion control. Surprisingly, neurons related to scanning eye movements differed from neurons activated during saccades suggesting functional specialization and segregation of both systems for eye movement control.     

Zoledronate Effects on Systemic and Jaw Osteopenias in Ovariectomized Periostin-Deficient Mice

Osteoporosis and periodontal disease (PD) are frequently associated in the elderly, both concurring to the loss of jaw alveolar bone and finally of teeth. Bisphosphonates improve alveolar bone loss but have also been associated with osteonecrosis of the jaw (ONJ), particularly using oncological doses of zoledronate. The effects and therapeutic margin of zoledronate on jaw bone therefore remain uncertain. We reappraised the efficacy and safety of Zoledronate (Zol) in ovariectomized (OVX) periostin (Postn)-deficient mice, a unique genetic model of systemic and jaw osteopenia. Compared to vehicle, Zol 1M (100 µg/kg/month) and Zol 1W (100 µg/kg/week) for 3 months both significantly improved femur BMD, trabecular bone volume on tissue volume (BV/TV) and cortical bone volume in both OVX Postn^+/+ and Postn^−/− (all p<0.01). Zol 1M and Zol 1W also improved jaw alveolar and basal BV/TV, although the highest dose (Zol 1W) was less efficient, particularly in Postn^−/−. Zol decreased osteoclast number and bone formation indices, i.e. MAR, MPm/BPm and BFR, independently in Postn^−/− and Postn^+/+, both in the long bones and in deep jaw alveolar bone, without differences between Zol doses. Zol 1M and Zol 1W did not reactivate inflammation nor increase fibrous tissue in the bone marrow of the jaw, whereas the distance between the root and the enamel of the incisor (DRI) remained high in Postn^−/− vs Postn^+/+ confirming latent inflammation and lack of crestal alveolar bone. Zol 1W and Zol 1M decreased osteocyte numbers in Postn^−/− and Postn^+/+ mandible, and Zol 1W increased the number of empty lacunae in Postn^−/−, however no areas of necrotic bone were observed. These results demonstrate that zoledronate improves jaw osteopenia and suggest that in Postn^−/− mice, zoledronate is not sufficient to induce bone necrosis.     

Meiotic Recombination Analyses in Pigs Carrying Different Balanced Structural Chromosomal Rearrangements

Correct pairing, synapsis and recombination between homologous chromosomes are essential for normal meiosis. All these events are strongly regulated, and our knowledge of the mechanisms involved in this regulation is increasing rapidly. Chromosomal rearrangements are known to disturb these processes. In the present paper, synapsis and recombination (number and distribution of MLH1 foci) were studied in three boars (Sus scrofa domestica) carrying different chromosomal rearrangements. One (T34he) was heterozygote for the t(3;4)(p1.3;q1.5) reciprocal translocation, one (T34ho) was homozygote for that translocation, while the third (T34Inv) was heterozygote for both the translocation and a pericentric inversion inv(4)(p1.4;q2.3). All three boars were normal for synapsis and sperm production. This particular situation allowed us to rigorously study the impact of rearrangements on recombination. Overall, the rearrangements induced only minor modifications of the number of MLH1 foci (per spermatocyte or per chromosome) and of the length of synaptonemal complexes for chromosomes 3 and 4. The distribution of MLH1 foci in T34he was comparable to that of the controls. Conversely, the distributions of MLH1 foci on chromosome 4 were strongly modified in boar T34Inv (lack of crossover in the heterosynaptic region of the quadrivalent, and crossover displaced to the chromosome extremities), and also in boar T34ho (two recombination peaks on the q-arms compared with one of higher magnitude in the controls). Analyses of boars T34he and T34Inv showed that the interference was propagated through the breakpoints. A different result was obtained for boar T34ho, in which the breakpoints (transition between SSC3 and SSC4 chromatin on the bivalents) seemed to alter the transmission of the interference signal. Our results suggest that the number of crossovers and crossover interference could be regulated by partially different mechanisms.     

The role of fecundity and sexual selection in the evolution of size and sexual size dimorphism in New World and Old World voles (Rodentia: Arvicolinae)

Evolutionary ecologists dating back to Darwin (1871) have sought to understand why males are larger than females in some species, and why females are the larger sex in others. Although the former is widespread in mammals, rodents and other small mammals usually exhibit low levels of sexual size dimorphism (SSD). Here, we investigate patterns of sexual dimorphism in 34 vole species belonging to the subfamily Arvicolinae in a phylogenetic comparative framework. We address the potential role of sexual selection and fecundity selection in creating sex differences in body size. No support was found for hyperallometric scaling of male body size to female body size. We observed a marginally significant relationship between SSD and the ratio of male to female home range size, with the latter being positively related to the level of intrasexual competition for mates. This suggests that sexual selection favours larger males. Interestingly, we also found that habitat type, but not mating system, constitutes a strong predictor of SSD. Species inhabiting open habitats – where males have extensive home ranges in order to gain access to as many females as possible – exhibit a higher mean dimorphism than species inhabiting closed habitats, where females show strong territoriality and an uniform distribution preventing males to adopt a territorial strategy for gaining copulations. Nonetheless, variation in the strength of sexual selection is not the only selective force shaping SSD in voles; we also found a positive association between female size and litter size across lineages. Assuming this relationship also exists within lineages (i.e. fecundity selection on female size), this suggests an additional role for variation in the strength of fecundity selection shaping interspecific differences in female size, and indirectly in SSD. Therefore our results suggest that different selective processes act on the sizes of males and females, but because larger size is favoured in both sexes, SSD is on average relatively small.