Brain-Score Which Artificial Neural Network for Object Recognition is most Brain-Like?, bioRxiv, 2018-09-05

The internal representations of early deep artificial neural networks (ANNs) were found to be remarkably similar to the internal neural representations measured experimentally in the primate brain. Here we ask, as deep ANNs have continued to evolve, are they becoming more or less brain-like? ANNs that are most functionally similar to the brain will contain mechanisms that are most like those used by the brain. We therefore developed Brain-Score – a composite of multiple neural and behavioral benchmarks that score any ANN on how similar it is to the brain’s mechanisms for core object recognition – and we deployed it to evaluate a wide range of state-of-the-art deep ANNs. Using this scoring system, we here report that (1) DenseNet-169, CORnet-S and ResNet-101 are the most brain-like ANNs. (2) There remains considerable variability in neural and behavioral responses that is not predicted by any ANN, suggesting that no ANN model has yet captured all the relevant mechanisms. (3) Extending prior work, we found that gains in ANN ImageNet performance led to gains on Brain-Score. However, correlation weakened at ≥ 70% top-1 ImageNet performance, suggesting that additional guidance from neuroscience is needed to make further advances in capturing brain mechanisms. (4) We uncovered smaller (i.e. less complex) ANNs that are more brain-like than many of the best-performing ImageNet models, which suggests the opportunity to simplify ANNs to better understand the ventral stream. The scoring system used here is far from complete. However, we propose that evaluating and tracking model-benchmark correspondences through a Brain-Score that is regularly updated with new brain data is an exciting opportunity experimental benchmarks can be used to guide machine network evolution, and machine networks are mechanistic hypotheses of the brain’s network and thus drive next experiments. To facilitate both of these, we release <jatsext-link xmlnsxlink=httpwww.w3.org1999xlink ext-link-type=uri xlinkhref=httpBrain-Score.org>Brain-Score.org<jatsext-link> a platform that hosts the neural and behavioral benchmarks, where ANNs for visual processing can be submitted to receive a Brain-Score and their rank relative to other models, and where new experimental data can be naturally incorporated.

biorxiv neuroscience 200-500-users 2018

Re-Evaluating One-step Generation of Mice Carrying Conditional Alleles by CRISPR-Cas9-Mediated Genome Editing Technology, bioRxiv, 2018-08-31

AbstractCRISPR-Cas9 gene editing technology has considerably facilitated the generation of mouse knockout alleles, relieving many of the cumbersome and time-consuming steps of traditional mouse embryonic stem cell technology. However, the generation of conditional knockout alleles remains an important challenge. An earlier study reported up to 16% efficiency in generating conditional knockout alleles in mice using 2 single guide RNAs (sgRNA) and 2 single-stranded oligonucleotides (ssODN) (2sgRNA-2ssODN). We re-evaluated this method from a large data set generated from a consortium consisting of 17 transgenic core facilities or laboratories or programs across the world. The dataset constituted 17,887 microinjected or electroporated zygotes and 1,718 live born mice, of which only 15 (0.87%) mice harbored 2 correct LoxP insertions in cis configuration indicating a very low efficiency of the method. To determine the factors required to successfully generate conditional alleles using the 2sgRNA-2ssODN approach, we performed a generalized linear regression model. We show that factors such as the concentration of the sgRNA, Cas9 protein or the distance between the placement of LoxP insertions were not predictive for the success of this technique. The major predictor affecting the method’s success was the probability of simultaneously inserting intact proximal and distal LoxP sequences, without the loss of the DNA segment between the two sgRNA cleavage sites. Our analysis of a large data set indicates that the 2sgRNA–2ssODN method generates a large number of undesired alleles (&gt;99%), and a very small number of desired alleles (&lt;1%) requiring, on average 1,192 zygotes.

biorxiv genetics 200-500-users 2018

 

Created with the audiences framework by Jedidiah Carlson

Powered by Hugo