The number of genes linked to rare Mendelian genetic diseases has been estimated around 4,000 out of a total of approximately 30,000 genes in human genome [1-3]. It is clearly an underestimate of genetic contributions to rare diseases, as there are nearly 6,000 to 13,000 more genes to be identified for their roles in disease pathology [4]. Of approximately 6,100 unique rare diseases curated in Orphanet (www.orpha.net), those of genetic origins could make up to as much as 72%. A recent report has estimated the point prevalence, representing the population burden, less than 1/1,000,000 for the majority of these rare diseases [5]. Based upon this estimation, the overall population prevalence of rare diseases ranges from 3.5% to nearly 6%, affecting as many as 446 million people worldwide [5], with neurological illnesses as the most prevalent category [6]. While relatively low in their prevalence in the general population, the psychological and social burden of these rare diseases cannot be easily ignored, considering the cost and effort to find a definite molecular diagnosis that could take 4.8 years or longer on average [7, 8]. Such a long journey, or so-called “the diagnostic odyssey”, can be further complicated with the uncertain nature of pathogenicity associated with genetic variations in individual patients. The lack of our knowledge in functional characteristics of these variations can be devastating, given that it is virtually impossible to develop therapeutic options without prior knowledge of their physiological consequences.

The current effort to tackle rare human diseases has been facilitated by the widely used genome sequencing technologies, including whole-genome and whole-exome sequencing. However, the rate of success in identifying genetic causes with this methodological approach is rather limited, reaching less than 30% among patients referred for diagnosis [9, 10]. Furthermore, the ultimate quest to the pathogenicity of individual genetic variants remains largely unresolved even in approximately 8% of cases with identified genes [11]. While recent advances in in silico platforms have helped us to predict the pathogenic profile of variants, the accuracy of their prediction can vary among different algorithms, with approximately 80% of agreement on individual pathogenicity and up to 11% of incorrect prediction [12]. As an alternative to in silico prediction paradigms, multiple putative variants with similar phenotypes associated can be assembled together, which may lead to identification of a gene responsible for a unique rare disease. However, such approach could be cost-ineffective with a relatively slower rate of diagnostic decisions and still require validation of the pathogenicity with functional analyses.

In this review, we provide a brief overview of invertebrate model organism (MO)-based approaches to overcome limitations of the current research efforts and to recapitulate the core values of two representative models, Caenorhabditis elegans and Drosophila melanogaster, in our understanding of rare neurobiological diseases. In addition, examples of recent advances in bioinformatical and network-based activities will be addressed in detail to describe how they are transforming research paradigms. Finally, with some challenges to invertebrate MOs, future perspectives of these MO-based research will be discussed to refine our strategies in ending a painful journey of patients who seek to find an answer for their undiagnosed neurological illnesses.

Invertebrate MOs have helped us to broaden our understanding of biological phenomena across phyla for decades, mostly attributed to the significant degree of conservation observed in molecular mechanisms underlying major biological processes essential for cellular functions. For instance, discoveries in Caenorhabditis elegans (C. elegans) have delineated basic principles of RNA interference [13] and apoptosis [14]. Similarly, the early findings in Drosophila melanogaster (D. melanogaster) have provided critical information about the innate immunity and organ growth regulation [15-17]. While a degree of sequence identity at the level of DNA may remain relatively low, the presence of interchangeable orthologs sharing the core functions among different species further strengthens the value of these invertebrate MO-based analyses of human diseases. In addition, a single gene in an invertebrate model often corresponds to multiple human or vertebrate counterparts, or “co-orthologs”. Such reduced redundancy in invertebrates could be beneficial in uncovering the biological significance of human gene products with multiple subtypes or isoforms when their roles in biology remain elusive. Another cumbersome bottleneck in vertebrate animal model studies often lies in their ineffectiveness in cost and time management. Simple invertebrate models such as C. elegans and D. melanogaster are advantageous in this respect with relatively short developmental time and reduced cost of breeding. With the ease of genetic manipulations of their genomes, these invertebrate models allow extensive exploration of their phenotypes in significantly reduced time spans. In addition, well-established tools for efficient phenotyping in these organisms are widely available to facilitate investigation of genotype-phenotype relationships crucial for understanding the pathophysiology of rare diseases. With these concepts in mind, the important features of two representative invertebrate models, C. elegans and D. melanogaster, will be illustrated below, along with some exemplary experimental approaches adopted in D. melanogaster in great detail.

The Caenorhabditis elegans model

C. elegans is the first MO with its genome fully sequenced in 1998 [18]. A recent genomic analysis has revealed a significant degree of similarity between human and C. elegans genomes, with more than 50% of genomes shared while apparent genetic simplicity reflected in relatively low numbers of paralogs [19]. As stated above, a short generation time of four days allows rapid functional characterization of potentially pathogenic variations discovered in the gene of interest [20]. Furthermore, recent advances in genetic manipulation using the CRISPR-Cas9 gene editing technology have facilitated precise manipulation of genes at an individual nucleotide level [21-24]. Another key feature integral to C. elegans-based study is well-established knowledgebase and resources that are publicly available, including WormBase (https://www.wormbase.org) and Caenorhabditis Genetics Center (https://cgc.umn.edu), to facilitate integration of novel discoveries with the existing knowledge of disease pathology.

Recent achievements in C. elegans study have indeed significantly contributed to our understanding of rare human diseases. For instance, a missense mutation in Sodium Leak Channel, Non-Selective (NALCN) identified in a single child was successfully modeled in C. elegans [25]. This child was clinically presented with intellectual disability and ataxia as well as congenital arthrogryposis. With the conserved residue of human NALCN (R1181Q) identified in two C. elegans homologs, nca-1 and nca-2 [26], a orthologous substitution of the conserved residue was introduced in nca-1, followed by phenotypic characterization of mutants. Importantly, the mutants harboring this variant displayed abnormal locomotive behaviors recapitulating the gain-of-function (GOF) phenotypes previously described in other nca-1 mutant variants, thus confirming the GOF pathogenicity of a human NALCN variant [25]. A successful implementation of C. elegans model in rare disease research has also been demonstrated with the help of CRISPR-Cas9 gene editing technology. Two missense variants of B9D2 (P74S and G155S) identified in patients with ciliopathies due to Joubert syndrome [27] were successfully modeled in C. elegans by gene editing in homologous mksr-2 with over 60% of sequence identity in gene products. Introduction of presumed pathogenic mutations in C. elegans resulted in severely disrupted ciliary function and structure [28]. This study further demonstrates the value of C. elegans model in deciphering the functional consequences of potentially pathogenic variations causative of rare human diseases.

The Drosophila melanogaster model

Since its early contribution by Thomas H. Morgan in 1910s [29, 30], D. melanogaster has been at the heart of numerous breakthroughs indispensable for our current understanding of biological phenomena. As in worms, flies have a short developmental time around 10~14 days as well as a relatively short life span generally not exceeding three months. Together with a short generation time, a large number of progenies are produced per each reproduction cycle, thus making it an attractive in vivo model to perform large-scale screening experiments. In addition, a rich body of knowledgebase and experimental resources are widely available to the research community, including a solid database (FlyBase, https://flybase.org), public stock centers (BDSC, https://bdsc.indiana.edu; VDRC, https://stockventer.vdrc.at), and a cDNA clone collection (DGRC, https://dgrc.bio.indiana.edu). With its genome completely sequenced in 2000 [31], the degree of conservation between human and fly genomes is significant for coding genes (~65%), with the similarity even greater than 80% for the disease-associated human genes [32, 33]. In addition to genome-wide similarity, highly sophisticated genetic tools are widely available to manipulate genes of interest in a spatio-temporally controlled manner, thus facilitating a functional analysis of human variants with unknown pathogenicity in appropriate cellular or molecular contexts [34-37].

Experimental strategies to investigate rare human diseases using D. melanogaster can be divided into four different categories in general. 1) First, an attempt can be made to humanize flies by replacement of a fly gene with a human ortholog or homolog in its wild-type (WT) or variant form. Once the fly loss-of-function (LOF) mutants are generated and characterized, their phenotypes following introduction of a human ortholog are analyzed in great detail to see if the human counterpart in either a WT or a variant form can functionally rescue the LOF mutants [30]. For instance, a recent report on modeling of Parkinson’s disease has successfully demonstrated that L-DOPA-induced dyskinesia often observed in human patients could be phenotypically studied in Drosophila larval models by introducing a specific human variant of α-synuclein (A53T) and by monitoring dyskinetic behaviors following an administration of a high concentration of L-DOPA [38]. 2) When a variation in the gene of interest occurs in a conversed region between human and fly genomes, the pathogenicity of a specific variant can be directly investigated in flies by introduction of orthologous variants to an endogenous fly gene via CRISPR-Cas9-based gene editing [39, 40]. The degrees of rescue are compared between LOF mutants harboring either a WT or a variant fly cDNA to analyze the pathogenicity of each variant. 3) If a rapid assessment of a rare variant is required, one can force expression of human or fly cDNAs in a tissue-specific manner in WT animals. This will create an environment in which both endogenous fly gene products and exogenous fly or human counterparts are co-expressed, thus providing an opportunity to reveal a unique property of a variant, including a dominant-negative effect. 4) Finally, functional characterization of fly LOF phenotypes can also provide a clue to understand human variants of certain categories that are likely linked to a complete loss of gene functions, including non-sense, frameshift and deletion mutations. Such mutations in vertebrate models may lead to the lethality of an affected animal, thus making it impossible to study the pathogenicity of each variant and relevant molecular mechanisms. However, a knock-out of fly orthologs in a tissue-specific manner may offer a unique opportunity to study functional characteristics of a gene of interest. Together with the aforementioned intrinsic advantages, versatile applications and adaptations of these experimental paradigms further strengthen the value of D. melanogaster model in studying rare human diseases.

In the post-genome era, discoveries of genetic variants in the human genome further emphasize the role of bioinformatical analyses in deciphering the pathogenicity of each variant associated with rare human diseases of a definitive genetic origin. In line with this idea, large-scale sequencing centers have been established to facilitate the effective identification of potentially pathogenic disease variants, applying whole-genome and whole-exome sequencing to individuals or cohorts with suspected Mendelian diseases. For instance, the Centers for Mendelian Genomics was established in 2012, with a financial support from NIH in US [1]. A recent report from this institution in collaboration with investigators in 36 countries included an analysis of over 18,000 samples representing approximately 1,050 Mendelian phenotypes, which led to identification of nearly 1,000 genes associated with disorders, including 375 genes that have never been mapped to human diseases [2]. Identification of candidate variants from such large-scale sequencing approaches often requires initial validation of their pathogenicity with bioinformatical and computational means. For this purpose, a number of in silico prediction algorithms have been developed, aiding to refine the scope of subsequent functional analyses using MOs. In addition, recently established platforms such as CADD (Combined Annotation Dependent Depletion) and REVEL (Rare Exome Variant Ensemble Learner) incorporate machine learning technologies with previously developed algorithms to improve the power of prediction [41, 42]. Briefly, REVEL was established by incorporating a total of 18 pathogenicity prediction scores (8 conservation scores and 10 functional scores) from 13 tools and by introducing additional training sessions with recently unraveled diseases and rare missense variants that have not been previously provided for initial prediction training. In this way, extracted REVEL ensemble scores could provide an improved prediction value for the pathogenicity of missense and rare variants [41]. Similarly, CADD integrates multiple metrics with diverse information types, such as conservation and the nature of variants as missense changes, thus providing quantitative scores (“C-scores”) for predicting the deleteriousness of diverse arrays of genomic variations and for prioritizing potentially causative variants in research and clinical activities (https://cadd.gs.washington.edu/) [42].

This step of pathogenicity prediction can be further aided with bioinformatical analyses of genetic and physical interaction databases that provide valuable information about interactions between a gene or protein of interest and others previously associated with a specific Mendelian disease. Databases generated from large-scale proteomics screens in diverse species are included in this type of analysis, along with text-mining strategies for relevant publications based in MO studies. Recently, a few bioinformatical tools have been developed to optimize such analysis, including STRING (Search Tool for Recurring Instances of Neighboring Genes) and MIST (Molecular Interaction Search Tool). STRING provides a database for established and predicted genetic and protein interactions based upon systematic co-expression patterns, shared signals across multiple genomes, inferred interaction knowledge from different organisms and text-mining analysis of the scientific literature [43]. Similarly, the MIST platform integrates previously identified genetic and protein interactions from humans and multiple MOs, including yeast, worm, fly, zebrafish, frog, rat and mouse, to predict interactions inferred from “interlogs”, i.e. interactions between orthologous genes or proteins in different organisms [44]. A more comprehensive analysis of both sequence- and interaction-based datasets could further facilitate research efforts to delineate the pathogenicity of a rare variant potentially causative of undiagnosed human disease phenotypes.

Aside from the aforementioned large-scale screening efforts, discoveries of pathogenic variants often stem from isolate cases first identified by clinicians in practice. A translational approach from this individual discovery to MO-based research usually begins with integration of curated information across organisms. This step requires extensive literature and database search to see if a specific gene or variant of interest has been previously implicated in human diseases. The components of this initial analysis include, but not limited to, 1) the presence of prior reports of similar disease phenotypes, 2) the allele frequency of a specific variant in the general vs. disease population, 3) the availability of orthologs or homologs of implicated human genes in MOs, 4) their functional characteristics and expression profiles, 5) the site of variants in the gene structure, i.e. whether it is located in functional domains or not, and 6) the conservation of amino acids affected in variants across phyla. Manual curation and analysis of such diverse information is a near-impossible task, considering the sheer number of databases subject to analysis, thus requiring development of efficient tools to integrate multiple human and MO databases. One such tool available is MARRVEL (Model organism Aggregated Resources for Rare Variant ExpLoration) [45-47] (see also Fig. 1 for its integration in MO-based research networks described below). The current scope of MARRVEL encloses information obtained from 115,000 control cohorts and 12.3 million variants, genotype-phenotype relationships described for 6.96 million cases, and over 20,000 Gene Ontology terms corresponding to nearly 236,000 pairs of human genes and MO orthologs or homologs (MARRVEL v2.0). The human genetic databases incorporated in the MARRVEL analysis include ExAc [48], Geno2MP (NHGRI/NHLBI University of Washington-Center for Mendelian Genomics (UW-CMG)), ClinVar [49], DGV [50], DECIPHER [51] and OMIM (https://omim.org/). In addition, multiple MO databases are incorporated in the process of information curation, including SGD [52], PomBase [53], WormBase [54], FlyBase [55], ZFIN [56], MGI [57] and RGD [58]. The MARRVEL platform allows users to identify orthologs of a human gene of interest in MOs using DRSC Integrative Ortholog Prediction Tool (DIOPT) [32]. Following identification of available orthologs, it also provides comparison data at the level of protein, specifically concerning amino acid sequence alignment, annotation of functional domains and the presence of conserved residues.

While bioinformatical toolkits clearly help scientists to narrow the scope of putative targets of their research, there are some cautions to be made in interpretating the data generated with these means. First of all, the possibility of generating pseudo-positive and pseudo-negative results from in silico prediction should not be ignored. Indeed, a recent report on such prediction tools raised an issue of limited validity of prediction especially with clinical datasets [59]. In the case of interaction-based extrapolation of databases, the spatio-temporal context of each interaction needs to be carefully considered for proper assembly of information. Furthermore, functional significance of amino acid substitutions at relatively less critical positions or domains should still not be ignored and awaits further investigation in functional studies. Finally, the general assumption in database-derived assembly of information lies in the accuracy of the previously curated knowledge of genotype-phenotype relationships available in public databases. Such assumption should be granted upon the quality control of individual experimental studies incorporated, thus requiring us to handle such information with caution.

Cost-ineffectiveness of conventional approaches to identify pathogenic variants as well as limitations of in silico prediction and relevant bioinformatical tools strongly urge incorporation of MO-based functional studies for affirmative validation of potentially causative variants in a large scale. However, the successful implementation of MO-based studies is only granted with well-coordinated and extensively collaborative activities among independent teams dedicated to analyses of different rare disease features, engaging clinicians, bioinformatics specialists and research scientists. Recent efforts in research fields indeed reflect adoption of such idea in collaborative network-based approaches. Here I briefly discuss the key features and achievements of two representative research networks established with a great emphasis on MO-based research activities.

Undiagnosed Disease Network

The Undiagnosed Disease Network (UDN, https://undiagnosed.hms.harvard.edu) was initiated from a NIH-funded project to establish a system that would integrate clinical and research activities in deciphering the molecular basis of rare human diseases [60, 61]. It consists of an interrelating network of centers and cores that encompass expert researches in clinical medicine, bioinformatics and MO-based genetics. The network includes 12 clinical sites, a single coordinating center, sequencing and metabolomics cores, and three Model Organisms Screening Centers (MOSCs) along with a central biorepository. The three MOSCs comprises C. elegans and D. melanogaster cores as well as two zebrafish cores. Together with the bioinformatics team embedded in its core, the MOSC plays a crucial role in facilitating rare human disease studies affiliated with UDN. The general workflow through UDN involves multi-step and interactive collaborations among clinicians, bioinformaticians and MO scientists. Briefly, research applications are submitted to the coordinating center, some of which are forwarded to clinical sites for further evaluation specifically for disease phenotypes and genetic variations of probands and their family members. Results of clinical evaluation are then forwarded back to its coordinating center as well as to MOSCs and sequencing and metabolomics cores for further studies. The candidate genes selected for MO-based studies are prioritized and assigned to appropriate MO cores for functional analyses. The research outcomes following this step are again communicated with experts in clinical sites (Fig. 1). As a result of this extensive network-based collaboration, a total of 2,184 participant applications were accepted for study, resulting in 528 cases diagnosed and 362 variants evaluated with MOs. Among the evaluated applications, a total of 133 genes were assigned for functional analyses in MOSC cores, some of which were directly linked to diagnostic advancement (https://undiagnosed.hms.harvard.edu).

D. melanogaster and C. elegans have had a significant contribution to the current success of UDN as two major cores of MOSCs in the network. As of January 2022, a total of 21 functional in vivo studies have been reported for rare neurological disorders with an affiliation of UDN (https://undiagnosed.hms.harvard.edu). Among these reports, 12 of them included a systematic analysis of rare genetic variants using flies and worms (Table 1), followed by 5 zebrafish-based and 6 mouse-based studies. Indeed, the first report demonstrating effective integration of MOSCs involved a collaborative effort among clinicians, bioinformatics experts and a MO research group studying D. melanogaster [62]. The network-based analysis of a genetic variant in this case was initiated from a report of a single child with a neurodevelopmental disorder carrying a de novo variant in Early B-cell Factor 3 (EBF3). Following an initial report of the variant, two more cases were identified, with the same amino acid of EBF3 gene product affected. These initial clinical and sequencing results led to a subsequent analysis in Drosophila. The function of a fly ortholog of human EBF3, knot, was characterized using fly LOF mutants as well as rescue lines with human cDNA in either a WT or a variant form. A failure of genetic rescue with human EBF3 carrying patient-derived missense mutations confirmed the pathogenicity of these de novo variants isolated from just three patients [62]. Together with other examples of coherent integration of MOs in the UDN-affiliated activities (Table 1), these results consistently portrait a crucial role of these two MOs in successful identification of pathogenic variants causative of rare and undiagnosed neurological diseases in the interactive network-based research environment. While initially established in US, the horizon of UDN has extended to include participants and researchers around the globe to facilitate international collaboration for rare human diseases, resulting in the formation of the Undiagnosed Disease Network International [63] (https://www.udninternational.org/).

Rare Diseases Models and Mechanisms

The Rare Diseases Models and Mechanisms (RDMM) network (http://www.rare-diseases-catalyst-network.ca/) was established in 2014 to facilitate collaborations between clinicians and MO scientists in Canada [75]. Unlike UDN described above, its core structure is committee-based to identify and catalyze the connections between clinicians and scientists and supplemented with the Canadian RDMM Registry for its data curation. The overall project flow starts with a submission of a “Connection Application” concerning novel variants of interest to the advisory committee, which then evaluates the proposal based on the following aspects: 1) the quality of genetic data for diseases, 2) the severity of a disease reported and the need for medical intervention to affected patients, 3) a possibility of developing therapeutic options, 4) the population burden on either a specific or the general population and 5) the novelty of a biological pathway implicated in a disease of interest [75]. Once the application is approved by the advisory committee, the next decision is made by the scientific advisory committee to match it to MO scientists enrolled in the Canadian RDMM Registry. If matched, experts in MO research are invited to submit a “Model Organism Proposal Application” to be approved by the committee as well as the clinician who submits the “Connection Application”. Following an approval of this proposal, MO scientists would receive a catalyst fund to initiate an immediate collaborative research (Fig. 2).

The scope of RDMM network encloses the vast majority of both clinical and research communities in Canada. For instance, the RDMM aims to engage the clinical community dealing with rare diseases through established projects, including FORGE Canada Consortium [76] and the Treatable Intellectual Disability Endeavor protocol in British Columbia (TIDE BC) [77]. In conjunction with this end, the network also engages the MO research community by offering genetic knowledgebase for MO-based research. As a part of this effort, a total of nearly 12,200 genes are classified into three categories. This dataset includes 1) tier 1 for the genes of which functions were directly investigated with MOs in previous studies, 2) tier 2 for those awaiting immediate investigation with MOs and 3) tier 3 genes with some degree of relevance inferred from their previous studies based upon Gene Ontology terms [75]. The RDMM network has received 135 Connection Applications, in addition to direct submissions of 116 understudied candidate genes suggested from FORGE, Care4Rare and UDN. Among these initial submissions, a total of 105 functional studies were funded, including 85 MO catalyst projects connecting clinicians and MO experts in mouse, fly, zebrafish, worm, yeast and some protozoa [75]. The contribution of invertebrate MOs to recent achievements of the RDMM network is significant in that 17 out of 105 studies are built around D. melanogaster and C. elegans systems [75]. Importantly, two recent reports for rare neurological diseases have exemplified successful implementation of these invertebrate models in providing functional data to support genotype-phenotype correlation [78] and to validate novel gene discoveries [25]. While initially based in Canada, the RDMM network continues to expand its horizon to the globe, leading to establishment of other regional networks with a similar structural design, including RDMM Europe (https://solve-rd.eu/rdmm-europe/), Australian Functional Genomics Network (https://www.functionalgenomics.org.au/) and J-RDMM in Japan (https://j-rdmm.org/).

When conducting functional experiments to study rare human diseases in MOs, it is possible that the phenotypes described in MOs may have no clear relationship to human phenotypes reported from individual patients. This class of complications can be further divided into two different scenarios. First, the phenotypes observed in MOs may not be fully rescued by introduction of human cDNA. In this case, human protein orthologous to MO counterparts may not be functional or cause lethality in the MO background. For instance, a human protein induced for rescue may not form a functional complex with endogenous proteins, or induce toxicity in a mode similar to GOF alleles. In addition, the effects of WT and variant forms of human gene products may not differ from each other on restoring endogenous functions of a MO gene in the LOF background. In such case, an alternative investigation in different MOs should be further sought to rule out the potential pathogenicity of variants.

It should also be noted that MOs carrying a variant of interest may display phenotypes that are relatively unrelated to human diseases, but still allowing to be quantitatively scored. These phenotypes, often referred to as “phenologs”, should be carefully examined, as they may provide valuable insights into the basic biology of human gene of interest albeit a lack of phenotypic homology. The significance of a “phenolog” is mostly based upon the general principle of conservation observed in key cellular signaling pathways. Even with nearly interchangeable components of a signaling pathway, disrupted endogenous activity of these components may manifest completely different scorable phenotypes among different species. For instance, mutations in Drosophila notch result in dysmorphic features in wings, legs and eyes while similar perturbations in human genome are mostly associated with congenital diseases non-homologous to Drosophila phenotypes [79]. The careful evaluation of “phenologs” can thus still provide an important biological clue to molecular links between rare human diseases and novel variants with no or less information available for their functional significance.

The pathophysiology of certain human diseases may require organ- or tissue-specific contexts for the experimental studies. In that case, functional alterations need to be modeled in human- or vertebrae-specific organs, thus prohibiting invertebrate MOs from maximizing its usage for intended studies. A zebrafish model can be an appropriate alternative as it possesses virtually all of the same organs as humans. In addition to this context, it is also possible that a human gene of interest may not have orthologs or homologs in invertebrate MOs. Therefore, it is not likely that these models present direct functional evidence to validate the pathogenicity of human variants. Nevertheless, the data obtained from invertebrate MOs can still provide basic knowledge about the function of novel variants or genes of interest in the context of conserved core signaling pathways, thus acting as in vivo indicators of perturbed cellular functions.

As evidenced by recent success stories from network-based approaches such as UDN and RDMM, efficient modeling and molecular diagnosis of rare neurological disorders requires well-coordinated collaborations among clinicians responsible for identification of patients carrying rare genetic variants, bioinformatics specialists analyzing multi-omics and literature-based datasets and research scientists specialized in different MOs. Therefore, recruitment of sufficient numbers of experts in each field as well as experienced research coordinators should be granted to facilitate a systematic analysis of currently available information, to design and employ solid experimental paradigms, and to collectively interpret the findings, in order to provide an appropriate molecular diagnosis for neurological disorders of an unknown nature.

With increasing numbers of genomic data available for analysis, an inter-database match has become a critical step in identification of patient pools with similar genotype-phenotype relationships. Besides, an initial clinical identification of rare human diseases often comes with just a single patient carrying an uncharacterized variant. Therefore, it is necessary to recruit more cases with similar genotype-phenotype relationships to enhance the statistical power of subsequent analyses. Several tools or platforms have been recently established, including Matchmaker Exchange (MME, https://www.matchmakerexchange.org/) [80-82] that incorporates multiple databases such as GeneMatcher [83], Phenome Central [84], DECIPHER [51, 85], MyGene2 [86], AGHA Patient Archive (https://mme.australiangenomics.org.au), PatientMatcher (https://github.com/Clinical-genolmics/patientmatcher/), seqr (https://seqr.broadinstitute.org/matchmaker/matchbox), RD-Connect GPAP (https://platform.rd-connect.eu/), and IRUD [87]. So far, MME provides the most extensive coverage of inter-database comparisons for genotype-phenotype relationships, employed in at least 40 reports of gene discoveries from nearly 190,000 participants or cases (https://www.matchmakerexchange.org/). With a genomic analysis of patients with undiagnosed neurological phenotypes becoming a widely available diagnostic tool in clinics, accurate assessment of inter-database information will be an indispensable prerequisite of MO-based approaches before initiation of model-based functional studies.

While speedy and effective analyses can be performed in invertebrate MOs, one can still take advantage of other vertebrate MOs, including zebrafish and mouse, to fill the gaps left behind. Interactive collaborative efforts with the mouse model-based research groups such as International Mouse Phenotyping Consortium (https://www.mousephenotype.org/) [88] and Knockout Mouse Phenotyping Program (KOMP2, https://commonfund.nih.gov/komp2) will further facilitate the molecular diagnosis of rare neurological disorders (Fig. 3). In case of context-specific modeling of human diseases, invertebrate MOs may not be ranked as the choice of studies, as it needs to re-create a patient-specific condition. This can be alternatively built with the use of patient-derived induced pluripotent stem cells and subsequent generation of context-specific cell types and organoids [89, 90] (Fig. 3).

The most of our efforts in studying rare genetic diseases with a Mendelian basis has been restricted to potentially pathogenic variants within the coding regions. However, with more frequent applications of whole-genome sequencing technology, it is likely that we discover a significant fraction of variants residing in non-coding regions, including microRNAs, long non-coding RNAs and repetitive DNA sequences. The degree of conservation tends to be significantly low in regulatory genomic regions between human and invertebrate MO genomes. In addition, potentially pathogenic variants occurring in these regions of human genome may cause noticeable changes in expression profiles of other genes. With limited degrees of conservation, invertebrate MOs can be alternatively used to analyze the effect of altered gene expression portraited in human patients on disease phenotypes [91].

With the help of technical advances, the rate of novel gene or variant discoveries responsible for rare neurological diseases have continued to increase in the last few decades. Recent reports have estimated that there are nearly 6,000 to 13,000 more genes to be studied for their roles in Mendelian genetic disorders, many of which functions likely affect the integrity of the human nervous system [4, 6]. Among these candidates, a significant fraction ranging from 4,000 to 10,000 is expected to be novel discoveries [4], thus still representing formidable tasks remaining ahead. With an increasing rate of genotype-based identification of novel variants linked to rare neurological diseases, concerted team efforts among clinicians, bioinformaticians and MO scientists will be in greater demand (Fig. 3). These expert communities providing interactive support to one another will serve as cornerstones to help patients with undiagnosed rare neurological disorders to end their pain-staking journey in finding answers for a molecular diagnosis and cures.

This work was supported by the Financial Supporting Project of Long-term Overseas Dispatch of PNU's Tenure-track Faculty, 2019 to Ji-Hye Lee.

Contributions of Drosophila melanogaster and Caenorhabditis elegans studies conducted through UDN to identification of pathogenic variants responsible for rare neurological disorders