RNA sequencing data integration reveals an miRNA interactome of osteoarthritis cartilage

2019, BMJ Annals of the Rheumatic Diseases, 133 citation (27 impact factor)

RNA sequencing data integration reveals an miRNA interactome of osteoarthritis cartilage

Abstract
내연골(cartilage) 에서 골관절염(OA, osteoarthritis) 병태생리학적 과정의 microRNA 상호작용을 밝힌다.
Mehods
데이터 : 환자와 건강한 관절 연골로부터 각각 35명씩 mRNA  시퀀스 데이터를 30명씩 miRNA 시퀀스 데이터를 얻어, 총 130 개 샘플을 확보
방법 : 
1. miRNA와 mRNA 에 대한 DE anlaysis 수행
2. miRNA와 mRNA 간 우선순위 지정 : inverse Pearson’s correlations 과 inverse DE of miRNAs and mRNAs 을 이용해
3. miRNA target 찾기 : 예측된(TargetScan/microT-CDS) 또는 실험적으로 검증된(miRTarBase/TarBase) 공개 데이터베이스에 대한 정보를 기반으로 필터링
4. Pathway enrichment
Result
step1 실행 결과, 142 DEmiRNAs, 2387 DEmRNAs
step2 & 3 실행 결과, 238개의 mRNA 표적을 갖는 62개의 miRNA에 대한 regulator network
step4 실행 결과, '신경계 발달(nervous system development)'과 관련된 유전자들이 miRNA 조절 메커니즘을 통해 조절될 가능성이 높다는 것을 밝혀냈습니다. 이 중 NTF3 유전자는 신경세포의 생존과 분화를 조절하는 역할을 하며, 신경성장인자와 밀접한 관련이 있습니다.

Conclusions
OA 연골의 miRNA와 mRNA 시퀀싱 데이터를 통합적으로 접근하여 OA miRNA 상호작용체와 관련 경로를 규명함

 

Introduction

 

- 관절염(OA)이 진행성 변화를 보이는 연골 및 골 아래연골과 관련한 노화 및 장애 관절 질환임

- epigenetic 메커니즘, 즉 miRNA와 같은 non-coding RNA의 병행 발현, DNA 메틸화 및 histone 수정 등은 생체내에서 발현 변화를 조절하는데, 이러한 메커니즘은 현재 chondrocyte와 같은 후생존세포에 특히 중요

- miRNA-mRNA 상호작용을 대상으로 한 치료법이 암 질환에서 miRNA 모방제 및 항miR 이 성공적으로 적용되어, 전임상 개발에 기여

- 연골의 miRNA 및 mRNA 의 유전자 발현 조절 통합 분석

 

 

Methods

Small RNA and mRNA sequencing

- detailed description on alignment, mapping and normalisation is available in online supplementary materials.

small RNA-seq analysis)

1. aligned to the GRCh37/hg19 reference human genome with the software Bowtie1 using best strata option.

2. Adapters were removed with Cutadapt v1.1 using 15 bp as a minimum length to keep after clipping.

3. Read count abundances was done with HTseq[3] and were further assigned to miRBase v21.[4]

 

RNA-seq analysis)

1. aligned using GSNAP[5] against GRCh37/hg19 using default parameters.

2. Read abundances per sample was estimated using HTSeq count.

3. Only uniquely mapping reads were used for estimating expression.

 

Quality control)

1. The quality of the raw reads from small RNA- and regular RNA-sequencing was checked using MultiQC.

2. The adaptors were clipped using Cutadapt v1.1 applying default settings (min overlap 3, min length).

3. To identify outliers, principal component analysis (PCA) was applied on both sequencing data sets (small RNA-seq and mRNA-seq). 

    - 데이터 차원을 축소하여 이상치를 찾아내기 위함

 

Differential expression analysis)

1. MicroRNAs and gene expression data were normalized and log2-transformed using the DESeq2 v1.20 R package.

2. Batch effect correction was performed using the function remove BatchEffect from the limma R package v 3.36.1

3. A general linear model (GLM) assuming a negative binomial distribution was applied followed by a paired Wald-test between preserved and lesioned OA cartilage samples.

    - 일반선형 모델, GLM : 선형 회귀 모델의 확장으로, 정규, 이항 포아송 등을 가정할 수 있다. 예측 변수와 결과 변수 사이의 관계를 모델링하는 데 사용

    - 여기서는 GLM 으로 음이항 분포를 사용 함, 포아송 분포의 일반화 버전

    - 두 집단에 대한 변수들을 음이항 분포로 모델링한 다음, 두 모델간의 차이를 검정하기 위해 paried Wald-test를 사용

4. Benjamini-Hochberg multiple testing corrected P-values with significance cut-off of 0.05 are reported as False Discovery Rate (FDR).

 

miRNA-mRNA network)

To generate an OA-specific miRNA-interactom

we applied anintegrated stepwise prioritization(= 우선순위화) approach to the identified set of DE miRNAs and mRNA based on:

1) significant negative Pearson correlation (|r| > 0.5 and p < 0.05) between miRNA and mRNA levels from 19 overlapping samples from 15 patients (4 paired = 8 paired sample and 11 non-paired samples);

2) difference in expression

(fold-change, FC)

of mRNA and miRNA between the paired samples being in

opposite direction(= miRNA 상향 발현되면 mRMA는 하향 발현되어야한다. 이는 supression 한다는 특징을 반영)

3) predicted miRNA-mRNA target pairs from TargetScan and microT-CDS using the default settings;

4) experimentally validated miRNA-mRNA target pairs downloaded from miRTarBase v7.0 and Tarbase v7

 

The resulting network was visualized using the RedeR (v 1.28) package.

 

Pathway enrichment)

1. Pathway enrichment analysis was performed using the online tool DAVID while selecting Gene Ontology terms for biological processes (GOTERM_BP_DIRECT).

2. Bonferroni multiple testing-corrected p values with a significance cut-off of 0.05 are reported as familywise error rate (FWER).

3. Enrichment analyses of the DE genes with FC ≥2 were performed separately for further comparison.

4. To specifically identify the miRNA regulated pathways in OA cartilage, enrichments of miRNA-target genes were performed using all significant DE genes (FDR<0.05) as background.

 

실행코드

https://git.lumc.nl/rcoutinhodealmeida/miRNAmRNA

Coutinhodealmeida / miRNAmRNA · GitLab

 

 

Results