2019-09-16
020_ Weighted-gene correlation network analysis of lithium response
Research Question and Aims
Response to lithium in Bipolar Disorder (BD) is both heritable and variable between individuals. This project investigates smallRNAome correlates of lithium response in 100 PsyCourse BD patients that suffer from BD I or II, and 277 control participants using weighted gene correlation network analysis (WGCNA).
Briefly, WGCNA is a systems biology approach that leverages pairwise correlation coefficients to reconstruct gene expression (or other 'omic data) networks. By hierarchically clustering gene expression data into modules of co-regulated genes and then assigning each module an"eigengene" (i.e. the first principal component of the expression profile of a given module) to enable association testing at the level of the module rather than individual transcripts, the multiple testing burden is significantly reduced. WGCNA is therefore more powerful than traditional transcriptome-wide approaches (e.g. differential expression analyses), normally resulting in dozens of candidate genes. Results can then be annotated and interpreted using e.g. interfaces to gene ontology software.
We have started to analyze lithium response using WGCNA before PsyCourse proposal were mandatory, our primary dependent variables are the continuous and the dichotomous lithium response phenotype as defined in PMID: 26806518.
Analytic Plan
A We will pre-process the collected small RNAome seq data from PsyCourse patients using standard and well-documented tools e.g. FastQC and Cutadapt (for quality and adapter trimming of reads). We will then apply miRDeep2 to produce count data for known miRNAs.
B We will run the above-mentioned miRNA count data through the WGCNA pipeline to identify relevant modules and candidate smallRNAs for lithium response, using only BD patients:
1. Construct a gene co-expression network
2. Identify modules
3. Relate modules to external information (here: both continuous and dichotomous lithium response phenotypes and covariates)
4. Study module relationships
5. Find key drivers in interesting modules
6. Annotation of results
C We may repeat the analyses described in A using control participants and good responders to lithium, to identify characteristics of this disorder subtype (see. e.g. PMID: 26503763).
D Depending on the result of the WGCNA analyses, key drivers in interesting modules may be further investigated using cell culture or mouse experiments.
Resources needed
Phenotype Data:
We have used the following variables from the PsyCourse3.0 dataset:
V1_id
V1_stat
v1_sex
v1_ageBL
v1_school
v1_prof_dgr
v1_ed_status
v1_curr_paid_empl
v1_disabl_pens
v1_spec_emp
v1_wrk_abs_pst_5_yrs
v1_cntr_brth
v1_cntr_brth_m
v1_cntr_brth_f
v1_cur_psy_trm
v1_outpat_psy_trm
v1_age_1st_out_trm
v1_daypat_inpat_trm
v1_age_1st_inpat_trm
v1_dur_illness v1_1st_ep
v1_tms_daypat_outpat_trm
v1_cat_daypat_outpat_trm
v1_adv v1_medchange
v1_lith v1_lith_prd
v1_fam_hist
v1_scid_dsm_dx
v1_scid_dsm_dx_cat
v1_scid_age_MDE
v1_scid_age_mania
v1_scid_no_mania
v1_scid_age_hypomania
v1_scid_no_hypomania
v1_scid_ever_halls
v1_scid_ever_delus
v1_scid_ever_psyc
v1_scid_yr_fst_psyc
v1_scid_evr_suic_ide
v1_scid_suic_ide
v1_scid_suic_thght_mth
v1_scid_suic_note_thgts
v1_suic_attmpt
v1_scid_no_suic_attmpt
v1_prep_suic_attp_ord
v1_suic_note_attmpt
v1_panss_X (all PANSS items of V1 and V4)
v1_idsc_itm_X (all IDSC-items)
v1_ymrs_X (all IDSC-items)
v1_cgi_s
v1_gaf
v4_lith
v4_lith_prd
v4_alda_A
v4_alda_B1
v4_alda_B2
v4_alda_B3
v4_alda_B4
v4_alda_B5
v4_alda_tot
Note that some variables were extracted from the original secuTrial phenotype database, because phenotype information from some individuals was not contained in the PsyCourse 3.0 dataset.
Biological data
We will use the smallRNAome seq data of the PsyCourse participants in question (clinical: 100 individuals, healthy controls: 277 individuals)