Deep neural networks (DNNs) are becoming increasingly pop-ular as a model of the human visual system. However, theyshow behaviours that are uncharacteristic of humans, includingthe ability to learn arbitrary data, such as images with pixel val-ues drawn randomly from a Gaussian distribution. We investi-gated whether this behaviour is due to the learning and memorycapacity of DNNs being too high for the training task. We re-duced the capacity of DNNs by incorporating biologically mo-tivated constraints – an information bottleneck, internal noiseand sigmoid activations – in order to diminish the learning ofarbitrary data, without significantly degrading performance onnatural images. Internal noise reliably produced the desiredbehaviour, while a bottleneck had limited impact. Combiningall three constraints yielded an even greater reduction in learn-ing capacity. Furthermore, we tested whether these constraintscontribute to a network’s ability to generalize by helping it de-velop more robust internal representations. However, none ofthe methods could consistently improve generalization.

Background

The past decade has witnessed remarkable progress towards resolution of the Tree of Life. However, despite the increased use of genomic scale datasets, some phylogenetic relationships remain difficult to resolve. Here we employ anchored phylogenomics to capture 107 nuclear loci in 29 species of acanthomorph teleost fishes, with 25 of these species sampled from the recently delimited clade Ovalentaria. Previous studies employing multilocus nuclear exon datasets have not been able to resolve the nodes at the base of the Ovalentaria tree with confidence. Here we test whether a phylogenomic approach will provide better support for these nodes, and if not, why this may be.

Results

After using a novel method to account for paralogous loci, we estimated phylogenies with maximum likelihood and species tree methods using DNA sequence alignments of over 80,000 base pairs. Several key relationships within Ovalentaria are well resolved, including 1) the sister taxon relationship between Cichlidae and Pholidichthys, 2) a clade containing blennies, grammas, clingfishes, and jawfishes, and 3) monophyly of Atherinomorpha (topminnows, flyingfishes, and silversides). However, many nodes in the phylogeny associated with the early diversification of Ovalentaria are poorly resolved in several analyses. Through the use of rarefaction curves we show that limited phylogenetic resolution among the earliest nodes in the Ovalentaria phylogeny does not appear to be due to a deficiency of data, as average global node support ceases to increase when only 1/3rd of the sampled loci are used in analyses. Instead this lack of resolution may be driven by model misspecification as a Bayesian mixed model analysis of the amino acid dataset provided good support for parts of the base of the Ovalentaria tree.

Conclusions

Although it does not appear that the limited phylogenetic resolution among the earliest nodes in the Ovalentaria phylogeny is due to a deficiency of data, it may be that both stochastic and systematic error resulting from model misspecification play a role in the poor resolution at the base of the Ovalentaria tree as a Bayesian approach was able to resolve some of the deeper nodes, where the other methods failed.

Multiple myeloma (MM) is a cancer of terminally differentiated plasma cells. MM remains incurable, but overall survival of patients has progressively increased over the past two decades largely due to novel agents such as proteasome inhibitors (PI) and the immunomodulatory agents. While these therapies are highly effective, MM patients can be de novo resistant and acquired resistance with prolonged treatment is inevitable. There is growing interest in early, accurate identification of responsive versus non-responsive patients; however, limited sample availability and need for rapid assays are limiting factors. Here, we test dry mass and volume as label-free biomarkers to monitor early response of MM cells to treatment with bortezomib, doxorubicin, and ultraviolet light. For the dry mass measurement, we use two types of phase-sensitive optical microscopy techniques: digital holographic tomography and computationally enhanced quantitative phase microscopy. We show that human MM cell lines (RPMI8226, MM.1S, KMS20, and AMO1) increase dry mass upon bortezomib treatment. This dry mass increase after bortezomib treatment occurs as early as 1 h for sensitive cells and 4 h for all tested cells. We further confirm this observation using primary multiple myeloma cells derived from patients and show that a correlation exists between increase in dry mass and sensitivity to bortezomib, supporting the use of dry mass as a biomarker. The volume measurement using Coulter counter shows a more complex behavior; RPMI8226 cells increase the volume at an early stage of apoptosis, but MM.1S cells show the volume decrease typically observed with apoptotic cells. Altogether, this cell study presents complex kinetics of dry mass and volume at an early stage of apoptosis, which may serve as a basis for the detection and treatment of MM cells.

Giant tortoises are among the longest-lived vertebrate animals and, as such, provide an excellent model to study traits like longevity and age-related diseases. However, genomic and molecular evolutionary information on giant tortoises is scarce. Here, we describe a global analysis of the genomes of Lonesome George-the iconic last member of Chelonoidis abingdonii-and the Aldabra giant tortoise (Aldabrachelys gigantea). Comparison of these genomes with those of related species, using both unsupervised and supervised analyses, led us to detect lineage-specific variants affecting DNA repair genes, inflammatory mediators and genes related to cancer development. Our study also hints at specific evolutionary strategies linked to increased lifespan, and expands our understanding of the genomic determinants of ageing. These new genome sequences also provide important resources to help the efforts for restoration of giant tortoise populations.

Female Aedes aegypti mosquitoes infect more than 400 million people each year with dangerous viral pathogens including dengue, yellow fever, Zika and chikungunya. Progress in understanding the biology of mosquitoes and developing the tools to fight them has been slowed by the lack of a high-quality genome assembly. Here we combine diverse technologies to produce the markedly improved, fully re-annotated AaegL5 genome assembly, and demonstrate how it accelerates mosquito science. We anchored physical and cytogenetic maps, doubled the number of known chemosensory ionotropic receptors that guide mosquitoes to human hosts and egg-laying sites, provided further insight into the size and composition of the sex-determining M locus, and revealed copy-number variation among glutathione S-transferase genes that are important for insecticide resistance. Using high-resolution quantitative trait locus and population genomic analyses, we mapped new candidates for dengue vector competence and insecticide resistance. AaegL5 will catalyse new biological insights and intervention strategies to fight this deadly disease vector.