Meta-Analysis of Parkinson's Disease Transcriptome Data Using TRAM Software: Whole Substantia Nigra Tissue and Single Dopamine Neuron Differential Gene Expression

The understanding of the genetic basis of the Parkinson's disease (PD) and the correlation between genotype and phenotype has revolutionized our knowledge about the pathogenetic mechanisms of neurodegeneration, opening up exciting new therapeutic and neuroprotective perspectives. Genomic knowledge of PD is still in its early stages and can provide a good start for studies of the molecular mechanisms that underlie the gene expression variations and the epigenetic mechanisms that may contribute to the complex and characteristic phenotype of PD. In this study we used the software TRAM (Transcriptome Mapper) to analyse publicly available microarray data of a total of 151 PD patients and 130 healthy controls substantia nigra (SN) samples, to identify chromosomal segments and gene loci differential expression. In particular, we separately analyzed PD patients and controls data from post-mortem snap-frozen SN whole tissue and from laser microdissected midbrain dopamine (DA) neurons, to better characterize the specific DA neuronal expression profile associated with the late-stage Parkinson's condition. The default "Map" mode analysis resulted in 10 significantly over/under-expressed segments, mapping on 8 different chromosomes for SN whole tissue and in 4 segments mapping on 4 different chromosomes for DA neurons. In conclusion, TRAM software allowed us to confirm the deregulation of some genomic regions and loci involved in key molecular pathways related to neurodegeneration, as well as to provide new insights about genes and non-coding RNA transcripts not yet associated with the disease.


Introduction
Filters for inclusion and exclusion of datasets in the analysis were applied as described in TRAM guidelines [20]. In particular, all the selected experiments were carried out on specific brain structure and clinical conditions (substantia nigra pars compacta from post-mortem brains or laser captured human dopaminergic neuron from individuals with PD and matched controls), based on availability of raw or pre-processed data.
Data from exon array or other probes were excluded, as a too high number of data rows could hinder the program execution. Other exclusion criteria were: the absence of identifiers corresponding to those found in the GEO sample records (GSM) or Array express sample records; platforms without standard format (for example with an atypical number of genes, i.e. <5.000 or >60.000); data whose expression values were not clearly identifiable as linear or logarithmic.
Searches were executed up to March 2015.

Literature Search
A systematic biomedical literature search was performed up to March 2015 in order to identify articles related to global gene expression profile experiments in PD patients. First, a general search using the common terms "Parkinson disease" and "microarray analysis" was carried out. Then, the MeSH terms "Parkinson disease", "microarray analysis" (or "gene expression profiling" or "oligonucleotide array sequence analysis"), "substantia nigra" and "human" were also used for a more advanced PubMed search.
All articles were cross-checked with database search results to find any additional available microarray data.
We used a version of TRAM including updated UniGene and Entrez Genes databases (TRAM 1.2, April 2014), in comparison with the original 2011 version [20].
For each series, we have downloaded the samples selected for the study in.txt format and subsequently they were divided in pools, based on the extraction method, in order to conduct the following analysis: TRAM SN ONLY, comparing the transcriptome map of whole substantia nigra of PD patient (pool A) and healthy control (pool B); TRAM DA ONLY, comparing the transcriptome map of laser microdissected DA neurons of PD patient (pool C) and healthy control (pool D).
In addition, the platforms not included in TRAM 1.2 version were manually extracted and imported. This step is required to associate the correct gene symbol to the probe identifiers in each experimental data set.
Finally, all samples grouped into folders for each pool, were imported in TRAM and automatically normalized by intra-and inter-sample normalization with default parameters [20]. Briefly, during expression data import all the gene or probe identifiers were converted to gene symbols via UniGene and then gene expression values were assigned to individual loci. According to the TRAM Guide available within the software, the intra-sample normalization was conducted with "Mean" default parameters, expressing each value as the percentage of the corresponding sample mean value. Likewise, the inter-sample normalization was conducted with "Scaled-Q" default parameters, a variant of quantile normalization useful to normalize data from platforms with highly different numbers of investigated genes [20].
For each locus, in each biological condition, TRAM calculated the expression value as the mean of all available values for that locus. The statistical significance was calculated taking into account all genes in the genome (genome median), in order to determine percentile thresholds to select over/under-expressed genes.
For both studies (TRAM SN ONLY and TRAM DA ONLY) we run the standard analysis "Map" mode, using default and single gene level parameters [20,23].
In "Map" mode with default parameters, TRAM searched for over/under-expressed segments which have a window size of 500,000 bp and a shift of 250,000 bp, defining a segment as over/under-expressed in a significative manner if the expression value was different between the two conditions and contained at least 3 over/under-expressed genes (genes at the top/bottom 2.5% of values). In "Map" mode with single gene level parameters, the window size was set to12,500 bp with a shift of 6,250 bp, which corresponds to about a quarter of the mean lenght of a gene. In this way the significant over/under-expression of a segment corresponds in most cases to that of a single gene.
The software used in our study should assess the possible risk of bias, as it is intrinsically resistant to the systematic differences between batches (groups) of samples, as previously described [20].

Database and literature search
GEO and Array Express wide search resulted in 71 series of expression data. We then filtered our analysis with criteria "substantia nigra" as specific area of interest and inclusion-exclusion restrictions (see Material and Methods section), achieving a total of 11 series. An additional data set was retrieved by the advanced PubMed search and kindly provided after correspondence with the author [16], see Table 1.
According to RNA source, 9 series were included in the TRAM SN ONLY analysis and 3 series were included in the TRAM DA ONLY analysis (Fig 1 and Table 1).
The total number of retrieved samples was 151 patients with symptomatic Parkinson's and subclinical disease (iLBD: incidental Lewy Bodies Disease [31]), and 130 samples as agematched controls. According to the different tissue collection (snap-frozen or laser capture microdissection) they were further subdivided in: pool A (PD patients SN ONLY); pool B (controls SN ONLY); pool C (PD patients DA ONLY); pool D (controls DA ONLY).
A complete description of sample identifiers and main sample features are listed in Table 1 and S1 Table. The selection strategy for eligible data is summarized in Fig 1.

Snap-frozen SN tissue PD patients vs. controls
We first analyzed differential expression of pool A (123 PD substantia nigra samples) versus pool B (104 control substantia nigra samples), derived from 9 series of expression data.
A total of 3,166,787 data points (gene expression value) from the pool A and 2,643,992 data points from the pool B, relative to 36,446 distinct loci for which an A/B ratio value was determinable, were included in the analysis (S2 Table).
"Map" mode analysis of pool A vs. pool B data resulted in 10 significantly over/underexpressed segments, mapping on 8 different chromosomes (Table 2, SN ONLY).
The higher expression ratio between PD patients and controls substantia nigra whole tissue was observed in chromosome 1 [coordinates 22,250,750,000], with three known genes of the human complement subcomponent C1q showing values within the higher 2.5th percentile: C1QA and C1QB and C1QC. Then TRAM analysis retrieved a segment on chromosome 7 [spanned from coordinates 23,250,001-23,750,000], containing two sequences referred to EST clusters, Hs.644466 and Hs.743502 and the FAM221A gene, encodes for a protein with unknown function. Analysis conducted on UniGene and EBI database indicate that FAM221A is highly conserved from human to zebrafish, with a specific expression in brain tissue. In the same segment, it is over-expressed also the known gene GPNMB, encoding for a type I transmembrane glycoprotein.
The third over-expressed segment was on chromosome 3 [coordinates 114,250,001-114,750,000], with the known gene TIGIT, encoding for a member of the PVR (poliovirus  Other three over-expressed segments were: on chromosome 1 [coordinates 89,000,001-89,500,000], including as statistically significant two known genes of the GBP family of guanylate binding protein, GBP3 and GBP1, and one EST cluster: Hs.732899; on chromosome 16 [coordinates 56,250,001-56,750,000], with several metallothionein superfamily genes marked as significative; on chromosome 13 [coordinates 24,750,001-25,250,000], with two known genes, PABPC3 and AMER2 encoding respectively for a poly(A) binding protein and for an APC membrane recruitment protein 2 and the EST cluster Hs.731897, exhibiting an altered expression. The first under-expressed segment, with the lowest expression value in PD patients, was located on chromosome 5 [coordinates 76,000,001-76,500,000] (Table 2), with two known genes that are characterized by a statistically significant under-expression in substantia nigra of PD patients: SV2C and IQGAP2, encoding respectively for synaptic vesicle glycoprotein 2C and a member of the IQGAP (IQ motif containing GTPase activating protein) family. The altered expression was also observed for the EST cluster Hs.646953. Then, there was the genomic segment on chromosome 7 [coordinates 65,750,001-66,250,000], containing one ncRNA (LOC441242) and two known loci, ASL and TPST1, encoding respectively for a member of the lyase 1 family and for the tyrosylprotein sulfotransferase 1. The third under-expressed segment spans the cluster of olfactory receptor genes located on chromosome 11 in the 11p15.4 region [coordinates 4,500,001-5,000,000], with some isoforms marked as statistically significant:  OR52I2, OR51A7, OR51G1 and OR51A2. Finally, the last under-expressed segment, with the same expression ratio of the previous one, was on chromosome 17 [coordinates 9,500,001-10,000,000]. The statistical significance was obtained for three known genes: WDR16, encodes for a WD repeat-containing proteins, USP43 an ubiquitin specific peptidase 43 and RCVRN encoding for a member of the recoverin family of neuronal calcium sensors. The segment also includes an over-expressed gene GAS7, which encodes for a growth arrest-specific 7 protein.
At single gene level (TRAM segment window of 12,500 bp), the known gene DEFA3, defensin, alpha 3 neutrophil-specific protein (chr8), has the highest expression value (Table 3 and S3  Table), followed by two known gene, AZGP1 (chr7) and PCDH20 (chr13), encoding respectively for alpha-2-glycoprotein 1, zinc-binding, and for a member of the protocadherin gene family, a subfamily of the cadherin superfamily.
Between the under-expressed genes in PD patients tissue, a fold increase lower than 20 is observed for three genes, in particular OR8H3, another gene of the olfactory receptor family, TEX22 named testis expressed 22, a protein coding gene with unknown function to date, and the ncRNA PSMG3 -AS1. Single gene level analysis of SN whole tissue data generated a total of 11,775 significative loci (see S3 Table for the complete list of over/under-expressed genes), corresponding to 1,217 single transcripts with altered expression.

Laser microdissected DA neurons PD patients vs. controls
The TRAM analysis of the 3 series of laser microdissected tissue expression data, pool C (28 PD neurons samples) vs. pool D (26 control neurons samples), processed a total of 1,098,430 data points from the pool C and 1,035,561 data points from the pool D, relative to 25,795 distinct loci for which an C/D ratio value was determinable (S4 Table).
"Map" mode analysis of pool C vs. pool D data resulted in 4 segments with statistical significance, mapping on 4 different chromosomes ( Table 2, DA ONLY). Three of the four segments show an over-expression and only one includes genes that are under-expressed in PD patients DA neurons.
The higher expression ratio between PD patient and control neurons was observed in chromosome 2 [coordinates 172,750,001-173,250,000], with one ncRNA MLK7-AS1 and two EST clusters marked as statistically significant. Then the second most over-expressed segment was on chromosome 11 [coordinates 1,000,001-1,500,000] and spanned the cluster of mucin genes in the 11p15.5 region, with an over-expression observed only for MUC6 and MUC5AC genes. The last segment is on chromosome X [coordinates 48,000,001-48,500,000], with three known genes belonging to the family of highly homologous synovial sarcoma X (SSX) breakpoint proteins showed expression values within the higher 2.5th percentile: SSX1, SSX4 and SSX4B.
Finally, the only segment resulted under-expressed in PD patients DA neurons was on chromosome 15 [coordinates 92,750,001-93,250,000] ( Table 2), and included one known gene RGMA encoding for a glycosylphosphatidylinositol-anchored glycoprotein, the LOC100507217 locus encoding for a ncRNA and the sequence named Hs.741028, which is to date uncharacterized.
At single gene level (TRAM segment window of 12,500 bp), the long non-coding LINC00520 mapping on chromosome 14 has the highest expression value (Table 4 and S5 Table), followed by two EST clusters (Hs.512440, Hs.554217) mapping respectively on chromosome 8 and chromosome 20. Fold increase higher than 4 is observed also for MUC4 and MUC6 genes, located respectively on chromosome 3 and 11, belonging to the mucin family, and for LYG2 on chromosome 2, encoding for a protein with lysozyme activity. Then also the ncRNA, MLK7-AS1 and the 3 EST cluster: Hs.291993, Hs.618995 and Hs.630709, show a similar fold increase.
Among the most under-expressed genes in PD patients DA neurons there are several genes that show a fold decrease from 3 to almost 5, including the known genes ALDH1A1 (chr9),

Comparison with previously published data
We compared our results with the ones obtained in the individual works that were included in TRAM meta-analysis. We selected the main genes resulted as differentially expressed in the previously published microarray studies and verified their expression profile in our analysis (S6 Table).
In Table 5 genes from at least two independent single studies are listed, with the expression values obtained from our two differential maps. A general trend of over/under-expression consistent with data available in the literature is confirmed (see references indicated).

Discussion
In this work, we proposed a transcriptome analysis of human substantia nigra, the most affected brain structure in Parkinson's disease. In particular, we have investigated the different expression profiles of PD patients brain compared to age-matched controls, considering data from snap frozen whole tissue as well as from isolated DA neurons, in distinct analyses. The aim was to better characterize the specific DA neuronal profile compared to the whole tissue The  section, including a large amount of cells other than DA neurons, as astrocytes, microglia and oligodendroglia cells. Usually, microarray analyses on dissected tissue revealed a set of deregulated genes, which is in agreement with the evidence that not only the DA neurons, but also other cells within the substantia nigra and adjacent brain regions, are involved in Parkinson's disease pathology [32]. The meta-analysis was conducted with TRAM software, a tool that can integrate data from different microarray experiments, performed on different platforms, through a method of intra-and inter-sample normalization (scaled quantile normalization), intrinsically not affected by the systematic differences between groups of samples in microarray experiments [20]. The problem of the batch effect and the statistical validity of TRAM has been previously discussed and confirmed in different recent studies [23,33,34].
We processed a relevant number of samples, derived from 9 series of expression data from post-mortem dissected tissue and 3 series of data from LMD DA neurons and the program allowed us to identify over/under-expressed critical genome regions, by comparing PD patients whole tissue vs. matched controls (pool A vs. pool B) or isolated DA cells vs. matched controls (pool C vs. pool D).
TRAM identifies critical genomic regions and genes with significant differential expressions between two biological conditions. In particular, it can be noted that when comparing PD isolated DA neurons vs. controls only a few significantly over-or under-expressed genomic regions are retrieved, indicating similarity between the transcriptome maps of the two conditions. It is well known that loss of neurons is considered a physiological condition typical of brain aging and post-mortem evidence suggests that the PD dopaminergic pathways are especially vulnerable to the effects of aging [35]. More recently also Elstner et al. have focused that PD expression profile of DA neurons, dramatically and specifically change when compared to younger control group instead of age-matched controls [17].
On the other hand, TRAM analysis highlights those particular regions that may discriminate the disease and that could therefore be essential for the identification of novel molecular pathways contributing to the pathogenesis of PD. The When comparing PD whole substantia nigra to that of age-matched controls, more genomic segments are retrieved by TRAM default analysis as a probable consequence of the SN tissue cell heterogeneity, and results obtained with both "Map" mode and single gene level analyses showed no overlapping data between SN and DA neuron expression profiles (Tables 2-4).
Moreover, a prevalent up-regulated activation of gliosis/inflammation specific genes is evident in the first analysis, likely due to late stages of PD patients.
Our results indicate that the most expressed segment in SN ONLY analysis is located on chromosome 1 (1p36), including C1QA, C1QB and C1QC genes, encoding for major constituents of the human complement subcomponent C1q (A chain, B chain and C chain, respectively). This seems to be consistent with the main presence of microglia cells in the sample, being the only cells that expressed C1q in SN and other brain areas [36]. It is known, in fact, that activation of the complement system promotes the removal of pathogens and tissue damage products from the brain and may be involved in neuronal cell death in neurodegenerative diseases. Besides, this observation is supported by recent results obtained in a mouse model of PD, showing that C1q is up-regulated in the nigrostriatal system [36].  [16,19] mitochondrial proton transport (GO:0042776) ATP5L chr11 0.99 0.59 [16,19] mitochondrial proton transport (GO:0042776) ATP6V1D chr14 0,89 0.57 [19,30] proton transport (GO:0015992) BEX1 chrX 0.70 0.41 [14,19,29,30] up regulation of transcription factor (GO:0045944)  [29,30] transcription, DNA-templated (GO:0006351) The known genes confirmed in at least two independent single studies are reported (see references indicated). Likewise, the second identified over-expressed segment in PD SN samples contain the known gene GPNMB, encoding a transmembrane protein, whose homologous has been shown to be implicated in the regulation of immune/inflammatory responses and expressed in microglia and macrophages in rat neural tissues [37]. Recent genome-wide meta-analysis studies have highlighted the GPNMB locus (7p15) as a new potential PD risk candidate gene [38]. Moreover, Tanaka and coll., provided evidence that GPNMB could have a potential protective role in neurodegenerative disorders, in particular in Amyotrophic Lateral Sclerosis (ALS) [39], suggesting that the locus should be further investigated also in PD models and human postmortem tissue.
Other regions marked as over-expressed in affected SN tissue contain inflammation-related gene loci: TIGIT (3q13.31), recently investigated for its role in immune regulation, especially in cancer and other chronic disease [40]; GBP3 and GBP1 (1p22.2), already been reported as differentially expressed in post-mortem brain studies [41].
Significantly, it is well known that the chromosome 1 regions which resulted relevant in our meta-analysis, have been previously considered as a hotspot for Parkinson's disease genes [42,43], confirming the efficiency of TRAM software. Another region of the genome identified as a locus for PD susceptibility and marked as the most under-expressed one in TRAM SN whole tissue analysis, is on chromosome 5 (5q13.3) [44]. In particular, the segment includes the known gene SV2C, encoding for the synaptic vesicle glycoprotein 2C, involved in neurotransmitter transport and densely expressed in dopaminergic neurons in substantia nigra [45]. A study by Nowack and coll. [46], has shown that modest changes in SV2C expression, in either direction, can have a significant impact on synaptic function, while a specific research on the genetic basis for nicotine effect on Parkinson's disease, identifies SV2C gene as a putative PDassociated gene [47]. Besides, results from previous studies reported the under-expression of other synaptic vesicle proteins, SV2A and SV2B, consistently with our data (see Table 5).
Other neuronal function-related genes were also retrieved by the single gene level analysis (Table 3), confirming other previous published expression data (Table 5 and S6 Table). In particular, NPTX2 gene whose up-regulation in Parkinsonian SN was established in single microarray analysis [14,15,29], and validated by experiments indicating a localization for NPTX2 protein in Lewy bodies, Lewy neurites and some glial cells [48].
CBLN1 gene was instead reported under-expressed across microarray studies including our TRAM analysis [13,29,30]. It encodes a cerebellum-specific precursor protein, precerebellin, with similarity to the globular (non-collagen-like) domain of complement component C1qB and seems to be a candidate for homeostatic regulation of synapse formation and maintenance [49].
At single gene level, between the twenty most over-expressed genes in PD brain tissue vs. controls, we observed the presence of two alpha-defensin genes located on chromosome 8 (Table 3): DEFA3, with the highest expression ratio (4.93) and DEFA1, with a 2-fold higher expression in PD pathological tissue. Again, defensins are a family of antimicrobial and cytotoxic peptides thought to be involved in host defense, confirming the possible microglial activity in response to the inflammatory status typical of the Parkinson's disease. To date, however, investigations conducted on alpha-defensins showed their involvement in inflammation typical of several diseases, such as AD [50] or diabetes [51].
Conversely, the meta-analysis conducted on microarray data from LMD DA neurons, has evidenced only few chromosome segments and loci directly related to metabolic pathways which are known to be involved in PD (e.g. immune response, protein ubiquitination, apoptotic process). Instead, our research results have indicated the prevalence of neuronal function genes, transcription factor and regulatory elements, as differentially expressed (Tables 2 and 4).
It is noteworthy that the most over-expressed segment contains the ncRNA transcript, MLK7-AS1 (2q31-q32) and that between the twenty most variable genes in single gene level analysis, several non-coding transcripts and not-yet-characterized EST clusters have been found, including LINC00520, with an almost ten-fold higher expression in PD patient cells.
Long non-coding RNAs (lncRNAs) are well studied among the thousands non-coding eukaryotic RNAs that have been discovered so far. Their cellular action mechanisms are still largely unknown [52], even if, accumulating evidence suggests that in the nervous system, lncRNA functions may regulate brain evolution and neural development [53], while other results suggest their involvement in neurodegenerative diseases, and specifically PD [54].
Another putative novel regulation sequence is miR-622, indicated by TRAM analysis as under-expressed in DA neurons of PD patients. MicroRNAs (miRNAs) are endogenous, small non-coding RNAs that regulate gene expression by antisense complementarity to the 3'-UTR region of specific mRNAs [55]. It is well known that miRNAs may regulate diverse biological processes such as cell proliferation, apoptosis, stress resistance, stem cell maintenance and cell identity [56]. Recent studies show that miR-622 is associated with tumor metastatic capability in gastric cancers [57] and can suppress glioma invasion and migration by directly targeting activating transcription factor ATF2 [58], but no data have been released about a miR-622 neuromodulation activity. Our result could suggest a role of miRNAs in the maintenance of dopamine neurons, consistently with a previous study where the authors investigated the role of several miRNAs in the terminal differentiation, function, and survival of mammalian midbrain dopaminergic neurons [59]. The authors identified miR-133b as specifically expressed in DA neurons and reduced in midbrain tissue of PD patients, establishing that it regulates the cells maturation and function, within a negative feedback circuit that includes the paired-like homeodomain transcription factor Pitx3 [59]. Besides, the pro-inflammatory and suppressive role of the most studied neuroimmune miRNAs, miR-155 and miR-146a, has been recently reviewed together with other miRNAs implicated in the pathophysiology of acute and chronic CNS diseases [60]. For this reason, it could be interesting to further investigate miR-622 for a possible similar role in post-transcriptional regulation.
By contrast to SN expression profile, characteristic of neuroinflammation and neurodegeneration, DA neurons expression profile underlies their ability to survive during these chronic processes in Parkinson's disease.
In this context, several studies show that the adaptive stress responses stimulated in various human neurodegenerative diseases, including PD, can confer resistance to a subsequent neurotoxic challenge [61].
In particular, amongst the most over-expressed genes in patients DA neurons, we can note some known genes involved in essential neuronal functions and survival, as the maintenance of cytoskeleton integrity (TRIM54 and GAS2L3) and the neuronal plasticity (SHISA7), known to be damaged processes in PD [62,63].
Contrariwise, between the most under-expressed transcript in patients we can point out ALDH1A1, specifically expressed in normal DA neurons and consequently under-represented in PD patients, as shown in several expression studies (see Table 5 and S6 Table); similarly, AGTR1 gene whose trend is consistent with previous evidence showing that the total cellular AGTR1 levels are drastically reduced in surviving dopamine neurons of PD patients [64].
Overall, a general under-expression profile of the DA neuron associated genes in PD substantia nigra is confirmed also by TRAM analysis (e.g.: TH, SLC6A3, DDC, EN1, see S6 Table), compatible with the neuron loss, suggesting that the over-expressed genes could act as moderator of the under-expression of specific genes related to PD and thus contribute to a neurodegenerative-resistant phenotype.
A peculiar result has emerged in PD neurons analysis ( Table 2, DA ONLY), as results indicate an almost two-fold higher over-expression of a segment on chromosome 11 (11p15.5).
The same locus was already investigated in a previous linkage study of juvenile parkinsonism, even if a specific linkage was excluded in the 10 affected individuals considered [65].
The region contains some members of MUC gene family and the single gene level evaluation has also indicated two mucin gene loci as the most over-expressed in DA neurons of Parkinson affected (Table 3). It is well known that mucins are co-secreted with trefoil factor family (TFF)peptides in a large number of human mucous epithelia [66].
TFF peptides are typical secretory products of a variety of mucin-producing epithelial cells, constituents of mucus gels with a demonstrated anti-apoptotic effects, and a probable modulation in inflammatory processes. The protective effect may operate by organising the mucin layer which protects the mucosa from damage [67]. Interestingly, TFF peptides have also been found widely expressed in rat and mouse nervous central system and in minor amounts in the human brain [66].
Furthermore, Belovari et al. [68], has recently investigated the presence of TFF1 and TFF3 in the nervous system of developing mouse embryos, suggesting their probable involvement in complex processes of nervous system development and differentiation, and brain plasticity.
Notwithstanding this, further studies are necessary to confirm the up-regulation observed in PD patients, as to date very few data are available about MUC genes expression in brain.
In conclusion, this study offers a new approach for the regional analysis of gene expression in Parkinson's disease, by combining multiple data sets from independent studies.
The results of our integrated research globally confirm the deregulation of genes involved in general key cellular functions (mitochondrial energy metabolism, protein degradation, synaptic function) as well as survival mechanisms (immune system processes, response to stimulus) and provide new insights about loci not yet associated with the disease. Further studies are needed to investigate the detailed roles of some of the coding genes and ncRNA resulted in this study.
Supporting Information S1 Table. Samples selected for the meta-analysis of gene expression profiles of PD patients vs. healthy controls. All Sample and Platforms IDs are related to GEO database. Age at death: patient/control age in years; PMI (hours): post mortem interval before freezing (the time is indicated in hours); CTR/PD: control/Parkinson's disease patient sample; Source: brain tissue and method of extraction; SN: substantia nigra whole tissue; DA: dopaminergic neurons from laser microdissection; Array-Platform-Title: type of array platform used in the analysis; Value type: normalization method; Platform and Sample rows: platform and sample spots number; References and GEO experiment reference. Ã The serie is not deposited in GEO database [16].