DeepLUCIA: predicting tissue-specific chromatin loops using Deep Learning-based Universal Chromatin Interaction Annotator.

MOTIVATION: The importance of chromatin loops in gene regulation is broadly accepted. There are mainly two approaches to predict chromatin loops: transcription factor (TF) binding-dependent approach and genomic variation-based approach. However, neither of these approaches provides an adequate understanding of gene regulation in human tissues. To address this issue, we developed a deep learning-based chromatin loop prediction model called Deep Learning-based Universal Chromatin Interaction Annotator (DeepLUCIA). RESULTS: Although DeepLUCIA does not use TF binding profile data which previous TF binding-dependent methods critically rely on, its prediction accuracies are comparable to those of the previous TF binding-dependent methods. More importantly, DeepLUCIA enables the tissue-specific chromatin loop predictions from tissue-specific epigenomes that cannot be handled by genomic variation-based approach. We demonstrated the utility of the DeepLUCIA by predicting several novel target genes of SNPs identified in genome-wide association studies targeting Brugada syndrome, COVID-19 severity and age-related macular degeneration. Availability and implementation DeepLUCIA is freely available at https://github.com/bcbl-kaist/DeepLUCIA. SUPPLEMENTARY INFORMATION: Supplementary data are available at Bioinformatics online.

Whole-exome sequencing identifies mutations of KIF22 in spondyloepimetaphyseal dysplasia with joint laxity, leptodactylic type.

Spondyloepimetaphyseal dysplasia with joint laxity (SEMDJL), leptodactylic (lepto-SEMDJL) or Hall type, is an autosomal-dominant skeletal dysplasia manifesting with short stature, joint laxity with dislocation(s), limb malalignment, and spinal deformity. Its causative gene mutation has not yet been discovered. We captured and sequenced the exomes of eight affected individuals in six unrelated kindreds (three individuals in a family and five simplex individuals). Five novel sequence variants in KIF22, which encodes a member of the kinesin-like protein family, were identified in seven individuals. Sanger sequencing of KIF22 confirmed that c.443C>T (p.Pro148Ser) cosegregated with the phenotype in the affected individuals in the family; c.442C>T (p.Pro148Leu) or c.446G>A (p.Arg149Gln) was present in four of five simplex individuals, but was absent in unaffected individuals in their family and 505 normal cohorts. KIF22 mRNA was detected in human bone, cartilage, joint capsule, ligament, skin, and primary cultured chondrocytes. In silico analysis of KIF22 protein structure indicates that Pro148 and Arg149 are important in maintaining hydrogen bonds in the ATP binding and motor domains of KIF22. We conclude that these mutations in KIF22 cause lepto-SEMDJL.

Destabilization and mislocalization of POU3F4 by C-terminal frameshift truncation and extension mutation.

Most X-linked nonsyndromic hearing loss is caused by various types of mutations of the POU domain class 3 transcription factor 4 gene (POU3F4). We found five unique missense and frameshift truncation and extension mutations in Korean patients. Two missense mutations (p.Thr211Met and p.Gln229Arg) disturbed transcriptional activity. Two frameshift extension mutations (p.Thr354GlnfsX115 and p.X362ArgextX113) were located outside of POU domain and nuclear localization signal (NLS) at the C-terminus. POU3F4 protein levels were low and could be restored by MG132, a proteasome inhibitor, in vitro. These mutant POU3F4 proteins were exclusively localized to the cytoplasm and did not have transcriptional activity. Frameshift mutation (p.Leu317PhefsX12) in POU3F4 leads to the truncation of the C-terminal 44 amino acids spanning the POU domain and NLS. This frameshift truncation mutant protein was located in both the nucleus and cytoplasm and was present at low protein levels. This mutant was also transcriptionally inactive, even in the presence of MG132. From these results, we conclude that frameshift truncation and extension mutations in the C-terminus of POU3F4 lead to cytoplasmic localization and subsequent proteosomal degradation due to structural aberrations, which cause transcriptional inactivity and thus nonsyndromic hearing loss.

Prediction of binding property of RNA-binding proteins using multi-sized filters and multi-modal deep convolutional neural network.

RNA-binding proteins (RBPs) are important in gene expression regulations by post-transcriptional control of RNAs and immune system development and its function. Due to the help of sequencing technology, numerous RNA sequences are newly discovered without knowing their binding partner RBPs. Therefore, demands for accurate prediction method for RBP binding sites are increasing. There are many attempts for RBP binding site predictions using various machine-learning techniques combined with various RNA features. In this work, we present a new deep convolution neural network model trained on CLIP-seq datasets using multi-sized filters and multi-modal features to predict the binding property of RBPs. With this model, we integrated sequence and structure information to extract sequence motifs, structure motifs, and combined motifs at the same time. The RBP binding site prediction on RBP-24 dataset was compared with two multi-modal methods, GraphProt and Deepnet-rbp, using area under curve (AUC) of receiver-operating characteristics (ROC). Our method (average AUC = 0.920) outperformed 20 RBPs with GraphProt (average AUC = 0.888) and 15 RBP with Deepnet-rbp (average AUC = 0.902). The improvement was achieved by using multi-sized convolution filters, where average relative error reduction was 17%. By introducing new RNA structure representation, structure probability matrix, average relative error was reduced by 3% when compared to one-hot encoded secondary structure representation. Interestingly, structure probability matrix was more effective on ALKBH5, where relative error reduction was 30%. We developed new sequence motif enrichment method, which we stated as response enrichment method. We successfully enriched sequence motif for 12 RBPs, which had high resemblance with other literature evidences, RBPgroup and CISBP-RNA. Finally by analyzing these results altogether, we found intricate interplay between sequence motif and structure motif, which agreed with other researches.

FARS2 mutation and epilepsy: Possible link with early-onset epileptic encephalopathy.

Early-onset epileptic encephalopathy (EOEE) consists of a heterogeneous group of epilepsy phenotypes. Recent technological advances in molecular biology have also rapidly expanded the genotype of EOEE. Genes involved in diverse molecular pathways, including ion channels, synaptic structure, transcription regulation, and cellular growth, have been implicated in EOEE. Mitochondrial aminoacyl tRNA synthetase, which plays a key role in mitochondrial protein synthesis by attaching 20 different amino acids to the tRNA tail, has been recently linked with the epilepsy phenotype. Here, we report a novel homozygous c.925G>A (G309S) missense mutation in the gene that encodes the human mitochondrial phenylalanyl-tRNA synthetase (FARS2) in four patients from two nonconsanguineous Korean families. All four patients suffered from intractable seizures that started at the age of 3 and 4 months. Seizure types were variable, including infantile spasms and myoclonic seizures, and often prolonged. Although their initial development seemed to be normal, relentless regression after seizure onset occurred in all patients. An etiologic investigation, including brain imaging and metabolic studies, did not reveal a specific etiology. We reviewed the epilepsy phenotypes of six additional FARS2 mutation-positive patients and suggest that FARS2 can be considered one of the genetic causes of EOEE.

A novel mutation of TMPRSS3 related to milder auditory phenotype in Korean postlingual deafness: a possible future implication for a personalized auditory rehabilitation.

UNLABELLED: Appropriate customized auditory rehabilitation for hearing impaired subjects requires prediction of residual hearing and progression of hearing loss. Mutations in TMPRSS3 encoding a transmembrane serine protease were reported to be associated with two different autosomal recessive nonsyndromic hearing loss (arNSHL) phenotypes, DFNB8 and DFNB10, in terms of residual hearing that may mandate different rehabilitation. We aimed to reveal the genetic contribution of TMPRSS3 mutations among Korean populations and to correlate the clinical phenotype with TMPRSS3 genotypes. Fifty families that segregated arNSHL and have visited our clinic recently for 2 years were recruited for TMPRSS3 screening. Novel TMPRSS3 variants detected in our cohort were modeled using a predicted three-dimensional (3D) structure of the serine protease domain. The prevalence reached up to 11.2 % (3/27) among subjects with either prelingual hearing loss but retaining some degree of language development or with postlingual ski-slope hearing loss. We also found that a p.A306T allele is a founder allele in this population. Based upon the 3D modeling, we were able to correlate significant retention of residual low-frequency hearing and slower progression of its loss to this novel variant p.T248M that was predicted to have milder pathogenicity. A yeast-based protease assay confirmed a mild pathogenic potential of the p.T248M variant and a tight correlation between the protease activity and the residual hearing. Preservation of this low-frequency hearing should be of utmost importance when considering auditory rehabilitation. Our results significantly narrow down the candidate population for TMPRSS3 sequencing for more efficient genetic diagnosis. More importantly, genotype-phenotype correlation of this gene observed in our cohort suggests that TMPRSS3 can be an appropriate candidate for personalized and customized auditory rehabilitation. KEY MESSAGE: The prevalence of TMPRSS3 mutations among Korean postlingual hearing loss is 8.3 %. The p.A306T variant of TMPRSS3 is the common founder allele in Koreans. A novel variant, p.T248M of TMPRSS3, was predicted to have milder pathogenicity. There was a genotype-phenotype correlation of this gene in Koreans. Our data support implication of this gene for personalized rehabilitation.

SH3 domain-peptide binding energy calculations based on structural ensemble and multiple peptide templates.

SH3 domains mediate signal transduction by recognizing short peptides. Understanding of the driving forces in peptide recognitions will help us to predict the binding specificity of the domain-peptide recognition and to understand the molecular interaction networks of cells. However, accurate calculation of the binding energy is a tough challenge. In this study, we propose three ideas for improving our ability to predict the binding energy between SH3 domains and peptides: (1) utilizing the structural ensembles sampled from a molecular dynamics simulation trajectory, (2) utilizing multiple peptide templates, and (3) optimizing the sequence-structure mapping. We tested these three ideas on ten previously studied SH3 domains for which SPOT analysis data were available. The results indicate that calculating binding energy using the structural ensemble was most effective, clearly increasing the prediction accuracy, while the second and third ideas tended to give better binding energy predictions. We applied our method to the five SH3 targets in DREAM4 Challenge and selected the best performing method.