Biology subjects

Xiao, H.-M.

Publications and source records attributed to Xiao, H.-M..

3 recordsLinked to original sources

bioRxiv2022

Multi-omics analyses of drug repurposing reveal Acebutolol and Amiloride for osteoporosis treatment

Osteoporosis is a metabolic bone disease that occurs during aging, characterized by low bone mineral density (BMD) and a high risk of trauma fracture. While current pharmacological interventions provide symptomatic benefits, they are unsatisfactory and have major side effects. In this study, we used multi-omics data and drug similarity to construct osteoporosis driver signaling networks (ODSN) and drug functional networks (DFN), respectively. By integrating ODSN and DFN with treatment transcriptional responses, we observed 8 drugs that demonstrated strong targeting effects on ODSN. Mendelian Randomization analysis determines the causal effect on BMD using cis-eQTLs of the drug targets and BMD GWAS data. The findings suggested Acebutolol and Amiloride may increase BMD, while Acenocoumarol, Aminocaproic acid and Armodafinil may enhance bone loss. Zebrafish experiments experimentally showed Acebutolol hydrochloride and Amiloride hydrochloride had significant protective effects on osteoporosis in zebrafish embryos induced by Dexamethasone. Also, Acenocoumarol reduced bone mineralization compared with the control group. The findings suggest that the hypertension drugs Acebutolol and Amiloride warrant further investigation in functional mechanistic experiments to evaluate their effectiveness for osteoporosis treatments.

systems biology↗

bioRxiv2021

Epigenomic and transcriptomic prioritization of candidate obesity-risk regulatory GWAS SNPs

Concern about rising rates of obesity has prompted searches for its genetic risk determinants in genome-wide association studies (GWAS). Most genetic variants that contribute to the increased risk of a given trait are probably regulatory single nucleotide polymorphisms (SNPs). However, identifying plausible regulatory SNPs is difficult because of their varied locations relative to their target gene and linkage disequilibrium, which makes most GWAS-derived SNPs only proxies for many fewer functional SNPs. We developed a systematic approach to prioritizing GWAS-derived obesity SNPs using detailed epigenomic and transcriptomic analysis in adipose tissue vs. heterologous tissues. From 50 obesity-related GWAS and 121,064 expanded SNPs, we prioritized 47 potential causal regulatory SNPs (Tier-1 SNPs) for 14 gene loci. A detailed examination of seven of these genes revealed that four (CABLES1, PC, PEMT, and FAM13A) had Tier-1 SNPs that might regulate alternative use of transcription start sites resulting in different polypeptides being generated or different amounts of an intronic microRNA gene being expressed. HOXA11 and long noncoding RNA gene RP11-392O17.1 had Tier-1 SNPs in their 3 or promoter region, respectively, and strong preferences for expression in subcutaneous vs. visceral adipose tissue. ZBED3-AS1 had two intragenic Tier-1 SNPs, each of which might contribute to mediating obesity risk through modulating long-distance chromatin interactions. We conclude that prioritization of regulatory SNP candidates should focus on their surrounding epigenetic features in a trait-relevant tissue. Our approach not only revealed especially credible novel regulatory SNPs, but also helped evaluate previously highlighted obesity GWAS SNPs that were candidates for transcription regulation.

bioinformatics↗

bioRxiv2020

Accurate recognition of colorectal cancer with semi-supervised deep learning on pathological images

BackgroundMachine-assisted recognition of colorectal cancer (CRC) has been mainly focused on supervised deep learning that suffers from a significant bottleneck of requiring massive labeled data. We hypothesize that semi-supervised deep learning leveraging a small amount of labeled data with abundant available unlabeled data can provide a powerful alternative strategy. MethodWe proposed a semi-supervised model based on the mean teacher architecture that provides pathological predictions at both patch- and patient-levels. We demonstrated the general utility of the model utilizing 13,111 CRC whole slide images from 8,803 subjects gathered from 13 independent centers. We compared our proposed method with the prevailing supervised learning and six pathologists. Two extended evaluations on 15,000 lung and 294,912 lymph node images were also performed to confirm the generality of utility of semi-supervised learning for different cancers. ResultsWith a small amount of labeled training patches ([~]3,150 labeled, [~]40,950 unlabeled or [~]6,300 labeled, [~]37,800 unlabeled), the semi-supervised learning (SSL) performed significantly better than the supervised learning (SL, which only used the labeled data) (area under the curve, AUC: 0.90 {+/-} 0.06 vs. 0.84 {+/-} 0.07, P value = 0.02 or AUC: 0.98 {+/-} 0.01 vs. 0.92 {+/-} 0.04, P value = 0.0004). Moreover, we found no significant difference between SL using massive [~]44,100 labeled patches and SSL ([~]6,300 labeled, [~]37,800 unlabeled) at patch-level diagnoses (AUC:0.98 {+/-} 0.01 vs. 0.987 {+/-} 0.01, P value = 0.134) and patient-level diagnoses (average AUC: 0.974 vs. 0.980, P value = 0.117). SSL was close to human pathologists in diagnosis performance (average AUC: 0.972 vs. 0.969). This extended evaluation on lung and lymph node also confirmed when a small amount of labeled data were used, SSL was better than SL, and achieved similar performance as that of SL with massive labeling. ConclusionsWe reported that SSL can achieve excellent performance through a multi-center study. Because SSL dramatically reduces the need and cost of pathological image annotation, it has great potential to effectively build pathological artificial intelligence (AI) platforms in practice.

pathology↗