Biology subjects

Gankin, Y.

Publications and source records attributed to Gankin, Y..

3 recordsLinked to original sources

bioRxiv2024

RNAfold: RNA tertiary structure prediction using variational autoencoder.

Understanding the three-dimensional (3D) organization of RNA is essential for advancing therapeutic development and vaccine design. However, the limited availability of experimentally resolved RNA structures restricts the applicability of data-intensive machine learning approaches for tertiary structure prediction. This study aims to develop a data-efficient method for learning coarse-grained RNA structural organization directly from sequence information. We propose AutoRNA, a variational autoencoder-based model that learns sequence-conditioned structural priors in the form of inter-nucleotide distance matrices. The model was trained on RNA structures obtained from the Protein Data Bank and restricted to sequences up to 64 nucleotides. Predicted distance matrices were converted into 3D coordinates using multidimensional scaling, followed by template-based assembly and molecular dynamics refinement. Model performance was evaluated using mean absolute error (MAE), root mean square error (RMSE), global distance test (GDT), and template modeling (TM) scores. On the test dataset, AutoRNA achieved an RMSE of 4.49~\AA{} and an MAE of 3.13~\AA{} for predicted inter-nucleotide centroid distances. However, the reconstructed three-dimensional structures showed limited global fold recovery, as indicated by moderate GDT scores and low TM-scores. Performance decreased for longer sequences, indicating limitations associated with data scarcity and increased structural complexity. Molecular dynamics refinement provided modest improvements, particularly for initially low-quality predictions. AutoRNA demonstrates that variational autoencoders can learn meaningful coarse-grained structural representations of RNA from limited data. While not suitable for near-native tertiary structure prediction, the method generates candidate coarse-grained inter-nucleotide distance restraints that could potentially be incorporated into downstream structure-reconstruction or physics-based refinement workflows. This work highlights the potential of generative models for RNA structure prediction in low-data regimes.

bioinformatics↗

bioRxiv2021

Analysis of 329,942 SARS-CoV-2 records retrieved from GISAID database

BackgroundThe 31st of December 2019 was when the World Health Organization received a report about an outbreak of pneumonia of unknown etiology in the Chinese city of Wuhan. The outbreak was the result of the novel virus labeled as SARS-CoV-2, which spread to about 220 countries and caused approximately 3,311,780 deaths, infecting more than 159,319,384 people by May 12th, of 2021. The virus caused a worldwide pandemic leading to panic, quarantines, and lockdowns - although none of its predecessors from the coronavirus family have ever achieved such a scale. The key to understanding the global success of SARS-CoV-2 is hidden in its genome. Materials and MethodsWe retrieved data for 329,942 SARS-CoV-2 records uploaded to the GISAID database from the beginning of the pandemic until the 8th of January 2021. To process the data, a Python variant detection script was developed, using pairwise2 from the BioPython library. Pandas, Matplotlib, and Seaborn, were applied to visualize the data. Genomic coordinates were obtained from the UCSC Genome Browser (https://genome.ucsc.edu/). Sequence alignments were performed for every gene separately. Genomes less than 26,000 nucleotides long were excluded from the research. Clustering was performed using HDBScan. ResultsHere, we addressed the genetic variability of SARS-CoV-2 using 329,942 worldwide samples. The analysis yielded 155 genome variations (SNPs and deletions) in more than 0.3% of the sequences. Nine common SNPs were present in more than 20% of the samples. Clustering results suggested that a proportion of people (2.46%) were infected with a distinct subtype of the B.1.1.7 variant. The subtype may be characterized by four to six additional mutations, with four being a more frequent option (G28881A, G28882A, and G28883[C] in the N gene, A23403G in S, A28095T in ORF8, G25437T in ORF3a). Two clusters were formed by mutations in the samples uploaded predominantly by Denmark and Australia, which may indicate the emergence of "Danish" and "Australian" variants. Five clusters were linked to increased/decreased age, shifted gender ratio, or both. According to a correlation coefficient matrix, 69 mutations correlate with at least one other mutation (correlation coefficient greater than 0.7). We also addressed the completeness of the GISAID database, where between 77% and 93% of the fields were either left blank or filled incorrectly. Metadata mining analysis has led to a hypothesis about gender inequality in medical care in certain countries. Finally, we found ORF6 and E as the most conserved genes (96.15% and 94.66% of the sequences totally match the reference, respectively), making them potential targets for vaccines and treatment. Our results indicate areas of the SARS-CoV-2 genome that researchers can focus on for further structural and functional analysis.

bioinformatics↗

bioRxiv2017

Performance Of Children With Autism In Parent-Administered Cognitive And Language Exercises

There is a broad scientific consensus that early and intensive therapy has the greatest chance of positive impact on an individual with autism spectrum disorder (ASD). However, the availability, quality, and general funding for early intervention programs is often lacking, leaving newly diagnosed children without adequate and sufficient therapy during the most critical early period of their development. Parent-administered iPad-assisted therapy has the potential to reduce the gap between the amount of therapy recommended for children with ASD and the amount they receive. However it is unclear how ASD severity and age influence a childs ability to engage with and learn from computerized cognitive exercises. In this manuscript, we describe data from a tablet-based therapeutic application administered by parents to 1,514 young children with ASD over the course of four to twelve months. We report that older children and children with milder forms of ASD performed better and progressed faster in cognitive and language exercises. However, most children were able to engage with and learn from exercises independent of their age or ASD severity. This data confirm that tablet-based cognitive and language exercises can be successfully administered by parents to children as young as two years of age over the course of many months independent of ASD severity.

clinical trials↗