Biology subjects

Akiyama, M.

Publications and source records attributed to Akiyama, M..

4 recordsLinked to original sources

bioRxiv2018

Direct inference of base-pairing probabilities with neural networks improves RNA secondary structure prediction with pseudoknots

MotivationExisting approaches for predicting RNA secondary structures depend on howto decompose a secondary structure into substructures, so-called the architecture, to define their parameter space. However, the architecture has not been sufficiently investigated especially for pseudoknotted secondary structures.\n\nResultsIn this paper, we propose a novel algorithm to directly infer base-pairing probabilities with neural networks that does not depend on the architecture of RNA secondary structures, followed by performing the maximum expected accuracy (MEA) based decoding algorithms; Nussinov-style decoding for pseudoknot-free structures, and IPknot-style decoding for pseudoknotted structures. To train the neural networks connected to each base-pair, we adopt a max-margin framework, called structured support vector machines (SSVM), as the output layer. Our benchmarks for predicting RNA secondary structures with and without pseudoknots show that our algorithm achieves the best prediction accuracy compared with existing methods.\n\nAvailabilityThe source code is available at https://github.com/keio-bioinformatics/neuralfold/.\n\nContactsatoken@bio.keio.ac.jp

bioinformatics↗

bioRxiv2018

Frequent variants in the Japanese population determine quasi-Mendelian inheritance of rare retinal ciliopathy

Hereditary retinal degenerations (HRDs) are Mendelian diseases characterized by progressive blindness and caused by ultra-rare mutations. In a genomic screen of 331 unrelated Japanese patients, we identify a disruptive Alu insertion and a nonsense variant (p.Arg1933*) in the ciliary gene RP1, neither of which are rare alleles in Japan. p.Arg1933* is almost polymorphic (frequency = 0.6%, amongst 12,000 individuals), does not cause disease in homozygosis or heterozygosis, and yet is significantly enriched in HRD patients (frequency = 2.1%, i.e. a 3.5-fold enrichment; p-value = 9.2x10-5). Familial co-segregation and association analyses show that p.Arg1933* can act as a Mendelian mutation, in trans with the Alu insertion, but might also cause disease in association with two alleles in the EYS gene in a non-Mendelian pattern of heredity. Our results suggest that rare conditions such as HRDs can be paradoxically determined by relatively common variants, following a quasi-Mendelian model linking monogenic and complex inheritance.

genomics↗

bioRxiv2017

A max-margin training of RNA secondary structure prediction integrated with the thermodynamic model

Motivation: A popular approach for predicting RNA secondary structure is the thermodynamic nearest neighbor model that finds a thermodynamically most stable secondary structure with the minimum free energy (MFE). For further improvement, an alternative approach that is based on machine learning techniques has been developed. The machine learning based approach can employ a fine-grained model that includes much richer feature representations with the ability to fit the training data. Although a machine learning based fine-grained model achieved extremely high performance in prediction accuracy, a possibility of the risk of overfitting for such model has been reported.\n\nResults: In this paper, we propose a novel algorithm for RNA secondary structure prediction that integrates the thermodynamic approach and the machine learning based weighted approach. Ourfine-grained model combines the experimentally determined thermodynamic parameters with a large number of scoring parameters for detailed contexts of features that are trained by the structured support vector machine (SSVM) with the{ell} 1 regularization to avoid overfitting. Our benchmark shows that our algorithm achieves the best prediction accuracy compared with existing methods, and heavy overfitting cannot be observed.\n\nAvailability: The implementation of our algorithm is available at https://github.com/keio-bioinformatics/mxfold.\n\nContact: satoken@bio.keio.ac.jp

bioinformatics↗

bioRxiv2017

Elucidating the genetic architecture of reproductive ageing in the Japanese population

Population studies over the past decade have successfully elucidated the genetic architecture of reproductive ageing. However, those studies were largely limited to European ancestries, restricting the generalizability of the findings and overlooking possible key genes poorly captured by common European genetic variation. Here, in up to 67,029 women of Japanese ancestry, we report 26 loci (all P<5x10{macron}8) for puberty timing or age at menopause, representing the first loci for reproductive ageing in any non-European population. Highlighted genes for menopause include GNRH1, which supports a primary, rather than passive, role for hypothalamic-pituitary GnRH signalling in the timing of menopause. For puberty timing, we demonstrate an aetiological role for receptor-like protein tyrosine phosphatases by combining evidence across population genetics and pre- and peri-pubertal changes in hypothalamic gene expression in rodent and primate models. Furthermore, our findings demonstrate widespread differences in allele frequencies and effect estimates between Japanese and European populations, highlighting the benefits and challenges of large-scale trans-ethnic approaches.

genetics↗