Systematic analysis of multi-omics data reveals component-specific blood-based biomarkers for Parkinson’s disease

Parkinson’s disease (PD) is a prevalent neurodegenerative disorder affecting millions of elderly individuals worldwide. Clinically, PD is diagnosed based on the presentation of motoric symptoms. Other methods such as F‑DOPA PET scan or α‑Synuclein detection from the cerebral spinal fluid are either too expensive or invasive for routine use. Omics platforms such as transcriptomics, proteomics, and metabolomics may identify PD biomarkers from blood, which can reduce cost and increase efficiency. However, there are many biological moieties being measured and issues with false positives/negatives. It is also unknown which omics platform offers most useful information. Therefore, it is important to assess the reliability of these omics studies. Here, we shortlisted and analysed nearly 80 published reports across transcriptomics, proteomics and metabolomics in search of overlapping blood‑based biomarkers for PD. The top biomarkers were reported across 29%, 42% and 12.5% of shortlisted papers in transcriptomics, prot‑ eomics and metabolomics respectively. These percentages increased to 42%, 60% and 50% accordingly when stud‑ ies were grouped by specific blood subtypes for analysis, demonstrating the need for test kits to be blood‑subtype specific. Following systematic analyses, we propose six novel PD biomarkers: two mRNAs (Whole blood, WB) – Arg1 and SNCA, two proteins (Plasma EV) – SNCA and APOA1, and two metabolites (WB) – 8‑OHdG and uric acid for fur‑ ther validation. While these proposed biomarkers are useful, they are also snapshots, representing subsets of larger pathways of origin where the different omics levels corroborate. Indeed, identifying the interconnections across differ‑ ent biological layers can strengthen contextual reasoning, which in turn, would give rise to better quality biomarkers. Knowledge integration across the omics spectrum revealed consistent aberrations on the same neuroinflammation pathway, showcasing the value of integrative (i)‑omics agreements for increasing confidence of biomarker selection. We believe that our findings could pave the way for identifying reproducible PD biomarkers, with potential for clinical deployment.


Graphical Abstract
Six Proposed blood-based biomarkers.Seventy-nine publications across transcriptomics, proteomics and metabolomics were shortlisted and analysed for reported biomarkers.The proposed biomarkers are SNCA, APOA1, Arg1, 8-OHdG and Uric acid.

Background: Parkinson's Disease and Current Diagnostic Tools
Parkinson's disease (PD) is a prevalent neurological disorder [1] characterized by the loss of dopaminergic neurons in the substantia nigra pars compacta of the midbrain [2].Since its discovery in 1817, the incidence rates of PD have increased by 10 times over the past nine decades [3].PD prevalence is strongly associated with the elderly, a phenomenon that aligns with the notion that age is a major risk factor for the development of PD [4].As the global population continues to age rapidly, PD poses a significant threat in deteriorating the quality of life for many afflicted individuals as well as their caregivers.PD is a progressive neurodegenerative disease typically diagnosed via neurological examination for classical motor symptoms such as bradykinesia and resting tremors [5].Clinically, PD diagnosis is usually conducted using the Unified Parkinson's Disease Rating Scale (UPDRS), which assesses an individuals' PD-related motor and non-motor deficits such as rigidity and olfactory dysfunction [6].Additionally, the UPDRS score is used to track disease progression, with increasing score indicating worsening of PD disability [7].However, even with clinical markers, the misdiagnosis rate for PD remains high at 42% [8,9].Clearly, less subjective phenotypic biomarkers would be helpful to improve PD diagnosis.To reduce subjectivity in PD diagnosis, brain imaging tools such as Magnetic resonance imaging (MRI) and Positron Emission Tomography (PET) have been used as a supplement and have shown potential to achieve a reliable diagnosis for parkinsonism.In particular, the 123I-ioflupane DaTSCAN was approved by the US FDA in 2011 for doctors to confirm a PD diagnosis [10].However, these imaging tools are expensive to conduct, and a negative result does not absolutely rule out PD [11,12].Therefore, a cheaper but feasible alternative is needed for PD diagnosis.
Recently, mounting evidence suggested that biofluids can reflect the pathophysiology of PD [13][14][15].This, in part, is fuelled by the Braak's hypothesis that sporadic PD begins via the olfactory or gastrointestinal system before affecting the central nervous system, suggesting that PD is not confined to the brain [16].Biofluids such as urine, blood, cerebrospinal fluid (CSF) and tear fluid have been studied for the metabolites of dopamine since PD is driven by the loss of dopaminergic neurons.CSF analysis could be a reliable prognostic tool as it better mirrors pathological changes of the brain [13,17].However, collecting CSFs is invasive and is not practically suitable in clinical settings for suspected cases of PD [13].Hence, this had led to the investigations of using non-invasive blood-based biomarkers such as serum, whole blood, and exosomes to facilitate PD diagnosis.With the rise of sequencing technologies, many studies have used omics to analyse diverse biological modalities of the genome, transcriptome, proteome, and metabolome in the blood of PD patients, with the view that they may provide valuable insights into the etiology of PD.However, current studies tend to be within the respective omics field and lacks cross-platform corroborations.This was demonstrated by Redenšek et al.where only 4% (5 out of 107 papers examined) of PD-related omics study from 2005 to 2017 were integrative [18].The lack of corroboration represents an important research gap.The overall contributory value of each individual study can be greatly enhanced by careful data mining and knowledge integration, to demonstrate how functionally coherent (or discordant) the targets reported across the omics spectrum are.Integrating multi-layered studies and identifying corroborations could reduce false-positive and false-negative results.It could also strengthen our contextual reasoning and understanding of the interconnections across different biological layers, thus deepening our insights into PD prognosis.To achieve this, we examined a total of 79 papers spanning transcriptomics, proteomics, and metabolomics, which revealed six blood-based biomarkers suitable for PD diagnosis.

Individual omics platform demonstrate diagnostic potential with improved agreement on biomarkers across studies for specific blood components
Advances in high-throughput "omics" technologies provides new ways of studying diseases.Omics refers to a field of biological study that has the suffix -omics, which include genomics, transcriptomics, proteomics, metabolomics and more recently microbiomics [19,20].Omics technologies provide insight not only at single biological moieties, but also at higher order functional structures such as biological mechanisms or pathways critical for initiating various diseases.For example, genome-wide association studies (GWAS) have identified common PD risk loci consisting of PARK16, ITPKB, MCCC1, SNCA, FAM47E-SCARB2, DLG2, LRRK2, RIT2 and FYN [21].Common risk loci like SNCA provides a handle for researchers to investigate the pathology of PD, for instance alternative splicing of SNCA risk loci can result in a SNCA112 transcript that results in SNCA proteins that are structurally more prone to aggregation [22].The aggregation of α-synuclein protein into Lewy bodies is the histopathological hallmark of PD [23].Through omics studies, we also now know that LRRK2 interacts with many other important proteins and play a central role in pathways underlying PD [24].
Individually, omics technology reveals some potential for identifying risk factors, diagnostic biomarkers and therapeutic targets for PD as summarized in Supplementary Table 1, 2, 3 and 4 across the 79 papers we examined.For each of the 18 transcriptomics-, 34 proteomics-, 27 metabolomics-related papers, we noted the list of reported targets and examined the agreement rate of reported targets across papers within each omics field.For each transcript, protein or metabolite, agreement rate is defined as the number of papers that reported it as biomarker divided by the total number of papers examined in the corresponding omics field.The highest agreement rate was 29% amongst papers publishing transcriptome signatures between 2007-2021, 42% amongst papers publishing proteome signatures between 2009-2021 and 12.5% amongst papers publishing metabolite signatures between 2009-2021 (Table 1).For each omic field, we further subdivided the transcript, protein or metabolite into the blood  subtypes (ie.Serum, plasma, EV etc.) that they were reported in and calculated the agreement rate within the blood subtypes.Importantly, the highest agreement rates increased to 42%, 60% and 50% for transcriptomics, proteomics and metabolomics respectively, when we grouped the papers by specific blood subtypes.This suggests that different blood subtypes reflect varying differential changes to PD that can be used as biomarkers and future studies should be specific about the component of blood being examined.

Multi-omics analysis suggests 6 blood component specific parkinson's disease biomarkers
Genetic analysis of biomarkers suggests the use of Arg1 and SNCA gene in whole blood as potential blood-based mRNA biomarkers of PD We included 15 studies that looked at gene upregulation and 17 studies that looked at gene downregulation across distinct blood subgroups (Table 2).Across all studies, commonly reported upregulated genes include Arginase 1 (ARG1) (26.67%) and Thrombomodulin (THBD) (26.7%).These genes were also reported in papers specific for whole blood.Interestingly, Arg1, an anti-inflammatory marker for anti-inflammatory microglia polarisation, is reported to be supressed in MPP + PD model [100] and affected by micro-RNA miR-155 in AAV2-SYN PD model [101].Given the role of neuroinflammation and microglia polarisation [102] in PD, we selected Arg1 as a candidate biomarker.On the other hand, the most downregulated gene across all papers is SNCA (29.4%).SNCA is also the most reported downregulated gene in papers on whole blood (42%).
Protein analysis of the various blood subtypes suggest use of SNCA and APOA1 in plasma EV as potential protein biomarkers of PD We identified 31 papers studying protein upregulation across various blood subtypes and 18 papers studying protein downregulation suitable for our intersection analysis (Table 3).The more commonly reported upregulated proteins across all blood subtypes in these papers are SNCA (42%), MAPT (10%), TTR (10%) and VWF (10%) (Table 3).When examining the specific blood subtypes, SNCA is still the most reported upregulated protein in plasma EV (60%), serum EV (57%) and plasma (25%).The more commonly reported downregulated proteins across all blood subtypes in the 18 papers are APOA1 (16.7%),FGG (16.7%),IGKV3-20 (16.7%) and SNCA (16.7%).Surprisingly, SNCA protein that is most reported as upregulated is also reported to be downregulated by other studies [53,55,73].Although it seems puzzling to observe downregulation in αSYN gene expression and upregulation in SNCA protein, various explanations have been proposed.For example, a downregulation in αSYN gene could be induced by upregulation in DNA methylation in the CpG sites that lead to the exhibition of PD phenotypes [103].Contrastingly, an upregulation of SNCA protein could be observed after autophagy-lysosomal pathway failure, where low-aggregated SNCA will predominantly be released via exosomes, in line with our observations of increased SNCA reported in plasma EVs Table 2 Analysis of top transcriptomic hits categorised by blood subtypes [104].Nonetheless, we selected SNCA as our biomarker of choice since its expression is reported by most as being affected in PD.We have also chosen APOA1 alongside SNCA as a downregulated biomarker for ease of testing using a common plasma EV blood subtype and also due to its relation to PD such as risk of having mild cognitive impairment as reported by other literatures [105][106][107].

Metabolites analysis of various blood subtypes suggests 8-OhDG markers for PD patients
We studied 26 and 24 papers on metabolites upregulation and downregulation in PD patients, respectively.
On the other hand, the more frequently reported upregulated metabolites across all studies are Proline (11.5%), 8-OhDG (7.69%) and Alanine (7.69%) (Table 4).8-OhDG was chosen as our choice of upregulated biomarker for ease of testing using a common whole blood subtype and also due to its relation to PD via oxidative stress as reported by others [108,109].

Overlap of biomarkers from multi-omics integration likely to be true signal amongst false positives or false negatives artefact from omics methodologies
Given that there are usually more expressed genes relative to acquired samples, transcriptomic analyses typically suffer from a lack of statistical power (curse-of-dimensionality) or produce many false positives/negatives due to erroneous assumptions on data distribution [110].To counter such issues, more appropriate bioinformatics algorithms were developed such as DESeq2 [111] or limma-voom [112].In addition, fluid-based proteomics also face many challenges and complexities.High abundance proteins such as albumin can mask low abundance proteins and must be removed to facilitate observation of lower abundance proteins.However, the removal of albumin might result in unintended removal of non-targeted low abundance proteins [113,114].Many peptides in serum also give highly intense signals that makes identification of endogenous peptides difficult.In addition, there is a wide array of technical variables that can influence the proteomic results [115].Variables such as blood withdrawal site or simply letting serum samples sit beyond 60 min can lead to detection of false targets arising from hemolysis caused by residual disinfecting alcohol [116] or unwanted cell lysis [117] respectively.The current way transcriptomics and proteomics are conducted and analyzed, may produce many false positives and false negatives.Multi-omics integration Table 4 Analysis of top metabolomic hits categorised by blood subtypes can value add by reducing the extent of false positives and false negatives presented.With the large amount of transcriptomics, proteomics, and metabolomics data available, targets that are repeatedly reported as common across the different omics layers are more likely to be true signals rather than technical artefacts.We define a cross-omics agreement rate as the intersection of transcriptomics and proteomics set of targets divided by the minimum number of targets of the 2 sets [ie.A ∩ B/min(A, B)].Analysis between the lists of compiled upregulated mRNAs and proteins identified 53 overlaps (13.7%) amongst the 2515 mRNAs and 386 proteins reported by the various studies.Similarly, between the lists of compiled downregulated mRNAs and proteins, there were 66 overlaps (12.9%) amongst the 1504 mRNAs and 508 proteins reported by the various studies.However, not all mRNA expression is positively correlated with protein expression [118].Hence, we examined cross-omics agreement regardless of direction of expression as well.There was an increase in percentage overlap with a total of 211 overlaps (23%) between the 4019 mRNAs and 894 proteins reported by the various studies.The Hypergeometric test is an important statistical instrument used to estimate the probability of chance occurrences of overlapping genes between 2 genes sets [119].Using the phyper() function in R and the assumption of 19,950 protein coding genes in the GRCh38 human genome [120], cross-omics biomarker overlap between all mRNAs and proteins reported is significant (pval = 0.005339753).This suggests that overlapping biomarkers via multi-omics integration are likely true signals and should be further validated, including our proposed biomarkers SNCA and Arg1.

Both transcriptomics and proteomics point to common theme of neuroinflammation and metabolic processes despite seemingly different list of biomarkers
Omics technologies can easily produce a large number of differentially expressed targets which is often overwhelming for researchers to look at individually.Hence, a common strategy is to identify the higher-order functional perspective by summarizing observed differential genes in light of their parent pathway (for instance, via DAVID) [121].Pathway analysis can improve experiment credibility by locating the most important pathways.For instance, when the protein list from each study was ran through EnrichR to obtain pathways affected for each study (Table 5), 33% reported downregulation of Complement and coagulation cascades for proteomics (vs 17% overlap in protein targets).Pathway analysis may also reduce the study's scope to a few hundred pathways rather than thousands of DEGs.Hence, in addition to intersections analysis, we exploited popular public repositories like Gene Ontology (GO) to study the pathways associated with PD.
We observed unifying themes for GO pathways derived from the compiled list of transcripts and proteins (Fig. 2).We pooled together upregulated differentially expressed genes from all the literature analysed and ran a pathway analysis using enrichGO function from clusterProfiler package in R (v4.0.5).The same was done for downregulated DEGs, upregulated proteins and downregulated proteins.Significant GO pathways from upregulated DEGs point towards IL6 and immune system changes, which were also observed in significant GO pathways from upregulated proteins.A study by Fielding et al. reported that IL6 is the key signal for neutrophil trafficking during inflammation, chemokine production and leukocyte apoptosis [122].Similarly, we observed significant pathways of IL6 production, regulation of IL6 production and positive regulation of IL6 production from upregulated DEGs, which were supported by significant pathways of neutrophil degranulation, neutrophil activation involved in immune response, cell killing and other inflammatory pathways involving MHC I from upregulated proteins.Despite seemingly modest overlap in the list of upregulated DEGs and proteins, PD is associated with neuroinflammation when evaluated collectively across omics layers as illustrated in the upset plot (Fig. 3A).The upset plot shows that different pathways were upregulated predominantly across the different omics, with neuroinflammation dominating transcriptomics; amino acid metabolism dominating metabolomics and blood related changes dominating proteomics (Fig. 3A).
Many significant pathways from downregulated DEGs and proteins are inflammation-related (eg.T-cell activation, lymphocyte proliferation and antigen processing / presentation).Other than neuroinflammation, the significant pathways from both downregulated DEGs and proteins converged on downregulated metabolic pathways such as cellular response to toxic substance, hydrogen peroxide catabolic process, ubiquitin-dependent protein catabolic process and regulation of cellular amino acid metabolic process.This reaffirms the proposition that PD is a metabolic disease [123] and also demonstrated in the upset plot, where the top hits other than neuroinflammation are protein and lipid metabolism (Fig. 3B).

Inter i-omics agreement on pathways can deepen confidence in selected biomarkers
The importance of integrating multi-layered studies is demonstrated when we observe how pathways across transcriptomics, proteomics and metabolomics corroborate.Pathways for the list of compiled metabolites     (Table 5) were obtained using MetaboAnalyst 5.0 [124].Arginine biosynthesis was the most frequent pathway, followed by biosynthesis of unsaturated fatty acids and ROS related pathways like aspartate and glutamate metabolism and glutathione metabolism.The metabolomic pathways exhibit great relevance to the transcriptomic and proteomic pathways identified.Arginine has been shown to inhibit acute microglia-mediated inflammation [125] while fatty acids can serve as inflammatory response signalling molecules [126].Additionally, glutathione metabolism is related to hydrogen peroxide catabolic process [127] that was picked up by proteomics.The interconnectedness across different biological layers comes to light with metabolomic pathways supporting the major neuroinflammation and metabolic themes highlighted by both transcriptomics and proteomics.Integration of cross omics findings can also result in greater confidence in differentiating targets with clinical value amongst a pool of false positives.For instance, knowledge integration of our proposed biomarkers SNCA, ARG1 and 8-OhDG reveals complex biological relationships.SNCA (α-syn) was shown to increase ARG1 in bone marrow derived macrophages [128], suggesting that SNCA may be responsible for triggering inflammation and immune response.In turn, there is observable increase in 8-OhDG, as interestingly, ARG1 also showed positive correlation with 8-OhDG levels [129].Hence, via i-omics agreement, there is increased confidence in our proposed biomarkers.

Pathway changes with age, motor severity and medication status agrees with disease progression
Information on the age range, UPDRS III assessment and medication status were used to subset the cohort into younger or older (> 67yrs.oldfor transcriptomics and proteomics, > 65 for metabolomics), less or more severe (> UPDRS mean) and not medicated or medicated for further pathway analysis.When segregated by age, transcriptomics of younger patient cohorts exhibited downregulation in detox and oxidative stress response pathways and upregulation of movement related pathways.Older patient cohorts subsequently exhibited upregulation of IL6 and inflammation pathways.In terms of proteomics, both young and old cohorts experienced dysregulation of various immune response.Younger cohorts also experienced downregulation of TCA cycle related metabolites.
We divided the cohorts obtained from studies that reported UPDRS III values based on their UPDRS scores (into low: < 22 for transcriptomics and proteomics, < 18 for metabolomics and high).At low UPDRS III scores, transcriptomic changes mainly involved downregulation of autophagy and upregulation of lipid metabolism related processes which was supported by metabolomic changes involving steroid hormone and fatty acid metabolism.Interestingly, proteomic changes at low UPDRS highlighted downregulation of many pathways involving dopamine uptake and biosynthesis as well as protein stability suggestive that changes start at low UPDRS status which leads to disease progression.Lastly, the transcriptomics of drug-naïve patient cohorts generally exhibited downregulated ROS processing and upregulated IL-6 production and inflammation.Downregulated phagocytosis but upregulated defence response to bacterium, killing of symbiont cells and immune responses were observed in proteomics.Metabolically, drug-naïve patient cohorts had downregulated energy production related to citrate cycle and pyruvate, but upregulated arginine and tryptophan metabolism related to inflammation as previously discussed.Interestingly, the drug treated patient cohorts showed some counteracting pathways.When treated, the transcriptomics of medicated patient cohorts had downregulated T cell activation and upregulated regulation of immune responses.Downregulated post translational protein modification and upregulated antigen processing, response to stress and protein removal were observed in proteomics.Metabolically, medicated patients had downregulated biosynthesis of unsaturated fatty acid and upregulated arginine biosynthesis.Taken together, medication had a positive effect on reducing inflammation and regulating protein stress.These pathway changes should however be taken with caution as they were not derived from paired studies of before and after treatment of the same patient but instead via comparison of different cohorts.
Taken together, by comparing age, motor severity and medication status, we observed that older, more severe and unmedicated patient cohorts exhibit dysregulation in energy [130], inflammation [131], lipid [132] and dopamine related pathways [133] which are in agreement with disease progression.A further examination of our 6 proposed biomarkers revealed preferences for different age, motor severity and medication status (Table 6), with the majority of the biomarkers favouring detection of younger PD patients.In agreement with our findings, previous studies have also identified reduced levels of APOA1 [134,135] and uric acid [136] to be associated with greater motor severity.

Standardisation of analysis and UPDRS patient staging can improve reproducibility and accuracy of biomarker identification
We observe modest overlap in agreement of targets across studies.However, this is unsurprising since the statistical thresholds, and multiple test corrections used in each independent study differed.We demonstrate this point by comparing the results produced via publicly available database and analytical tools on Gene Expression Omnibus (GEO)  repository and those reported by the original literature.We analyzed five microarray datasets to detect DEGs in PD patients, including GSE62283, GSE165083, GSE22491, GSE100054, GSE99039 datasets.Using the inbuilt GEO2R function on GEO repository and an intersection analysis (adjpvalue < 0.05, logFC > 1|logFC < -1) for the 5 datasets coded using R, we found that the most upregulated DEGs are CLTCL1, COMMD6, GNS, HGSNAT, LAMP2, LSM 3.00, LSMEM1, MANBA, SCARB2, SDPR, TCIRG1, and TPP1 as appeared twice in those microarray datasets (Table 7), while a lack of overlapping upregulated DEG is detected from the corresponding literature papers (Table 7).The full table of analysis from GEO2R can be found in Supplementary Table 5.Furthermore, the corresponding literature papers suggested BCL2 (using GSE6613 and GSE22491 datasets) [31,33] and TRAF6 (using GSE99039 and GSE22491 datasets) [30,31] are the most downregulated DEGs in PD patients, while a lack of overlapping downregulated DEG is identified using GEO2R analysis (Table 7).Interestingly, we found that XIST is more upregulated, and EIF1AY and KDM5D are more downregulated in females than in male PD patients using GSE7475 and GSE100054 datasets (Table 7), but this phenomenon has not been documented thus far.The lack of consistent overlapping DEGs between GEO2R analysis and the corresponding literature papers may be due to non-obvious confounding or batch factors that were known and corrected by the authors but not available in the inbuilt GEO2R function on GEO repository or due to our strict foldchange cut-off.In addition, the lack of overlapping genes seen in RNA sequencing data can also be attributed to the different analysis packages used (ie DESeq2 and edgeR).Dealing with these issues requires good quality meta-data.We therefore propose for more transparency in batch effects and standardisation of analysis pipelines by various studies to enhance reproducibility of results.Another observation was that studies do not properly synchronize and stage patient severity (or not published) via the universal UPDRS scale (~ 30% among our reviewed papers did not use UPDRS), which makes it difficult to properly find early diagnostic markers.In addition, some papers classified early PD as Hoehn and Yahr stage ≤ 2 [46], while other papers used Hoehn and Yahr stage 1 to 3 [14].This means that they are not comparing patients at the same time point and given how PD is a progressive disease, it is thus expected that different targets will be reported.Clearly, there is a need for better alignment between researchers and their clinician partners on how patient samples should be categorized and stored.A community-wide standardized framework for characterizing patient's PD status and disease severity using the UPDRS scale with a defined range of scores that distinguishes early PD from advanced PD patients would be useful.We also recommend that researchers use samples from the same PD staging and treatment to minimise confounding factors.This will aid in better comparisons and understanding of omics changes in PD progression.

Future directions: human-microbe i-omics
The gut microbiome is an exciting area to explore for PD given the emerging acceptance of a gut-brain axis in the pathogenesis of sporadic PD [16,137].Supporting the hypothesis, gut microbiome alterations have been reported by Toh et al., namely increased Akkermansia and reduced Roseburia in PD patients [138].In addition, Sampson et al. also demonstrated that gut microbiome from PD patients can induce enhanced motor deficits when introduced into germ-free mice [139].With the average 70 kg adult male hosting a total of 39 trillion bacteria in the body, humans are thus considered as supraorganisms and subjected to mutualistic microbiota-host interactions [140].Microbiomics hence hold great potential to identify new ways to screen risk factors, diagnostic factors and therapeutics to various human diseases [141] such as PD.Future research can focus on studying human-microbe interactions by integrating genomic, proteomic and metabolic analysis, which can lead to some novel and interesting insights.

Concluding remarks
We analysed 79 papers related to transcriptomics, proteomics and metabolomics profile of PD.Individual omics platforms revealed potential in identifying PD bloodbased biomarkers with increased reporting rates when grouped by blood subtypes, suggesting that different blood subtypes reflect different expression changes.Our study integrating the different omics datasets have validated SNCA as a potential useful biomarker and suggest Table 7 GEO2Ranalysis of top transcriptomic hits categorised by blood subtypes Arg1 expression, APOA1 protein level and two metabolites (WB) -8-OHdG and uric acid could add further value as additional biomarkers.It is also possible that some of these potential biomarkers are more sensitive or specific for certain PD subtypes (tremor dominant, gait disorder etc.) and these should be further validated in future studies.Corroboration of targets from different omics platforms can help overcome the false positives / negatives artefacts from current omics methodologies.Interestingly, when we pool all the genes and proteins together, there is indication of immunological and metabolic implications for pathways associated with PD.Neuroinflammation and metabolic disruption are common themes around the pathogenesis of PD, suggesting that despite the seemingly different biomarkers, i-omics agree in terms of overarching pathways.We also demonstrated how knowledge integration of cross omics findings show that our proposed targets SNCA, ARG1 and 8-OhDG work in the same pathway and increase our confidence in the proposed biomarkers.While current methodologies can yield biomarkers with potential for PD diagnosis, there can be greater transparency by authors in terms of the data cleaning and processing for greater reproducibility.Different data corrections used and thresholds selected for differential expression will naturally lead to different targets reported.Publicly available datasets from some of the 79 papers analysed using tools on GEO database resulted in different sets of targets reported by the literature.Lastly, we suggest that future studies standardise the usage of UPDRS scale and analyse samples from the same PD staging to minimise confounding factors for more accurate biomarkers.We also propose future research to make use of i-omics to gain deeper understanding of PD progression via human-microbe interactions related to the proposed the gut-brain axis in PD, the clarification of which holds promise to reveal prodromal markers for the disease.

Fig. 1
Fig. 1 PRISMA 2020 flow diagram for new systematic reviews.The screening process of 932 papers on PubMed related to Parkinson, Transcriptomics, Proteomics and Metabolomics to a final 79 papers for review that are blood-based specific

Fig. 2
Fig. 2 Gene Ontology (GO) Biological Processes (BP) Pathway Analysis indicates immunological and metabolic implications.A GO-BP pathways for compiled upregulated protein list.B GO -BP pathways for compiled upregulated transcript list.C GO-BP pathways for compiled downregulated protein list.D GO -BP pathways for compiled downregulated transcript list

Fig. 3
Fig. 3 Upset Plots Reveal Different Pathway Changes Dominated Different Source of Omics Analysis.A Upset plot for all pathways upregulated.Inflammation, amino acid metabolism and blood related changes dominated transcriptomics, metabolomics and proteomics respectively.B Upset plot for all pathways downregulated.Inflammation, Lipid metabolism and combination of Inflammation, Protein modification, Metabolic processes and polarity dominated transcriptomics, metabolomics and proteomics respectively

Table 5
Related pathways across transcriptomics, proteomics and metabolomics S No.

Table 6
Preference for age, UPDRS scoring and medication status of our proposed biomarkers.(-) indicates no information or no preference