Supplementary Figure 2: Over-represented GO terms in miR-140-3p-mediated regulatory network compared to the entire genome. (A) Dot plot for GO Biological Process terms. (B) Dot plot for GO Molecular Function terms. (C) Dot plot for GO Cellular Component terms. Each circle in the plots symbolizes an over-represented term: its x-axis coordinate reflects the Gene Ratio value; its size is directly proportional to the Count value; its color represents the Benjamini-Hochberg adjusted p-value generated by the hypergeometric test. Gene Ratio: the ratio of number of genes of interest that are annotated with a certain term from the database used to perform the analysis to number of genes of interest that are annotated with terms from the same database.

• Enciclopedia Taurina Cossio Pdf English

Count: number of node genes within the network that are annotated with a certain term. GO: Gene Ontology. Supplementary Figure 5: Over-represented GO Biological Process terms regarding nervous system and development in miR-140-3p-mediated regulatory network. Dot plot for all the most interesting GO Biological Process terms regarding nervous system and development. Each symbol in the plot symbolizes an over-represented term: its x-axis coordinate reflects the Gene Ratio value; its y-axis coordinate reflects the Background Ratio value; its color represents the Benjamini-Hochberg adjusted p-value generated by the hypergeometric test.

Gene Ratio: the ratio of number of genes of interest that are annotated with a certain term from the database used to perform the analysis to number of genes of interest that are annotated with terms from the same database. Background Ratio: the ratio of number of genes in the genome that are annotated with a certain term from the database used to perform the analysis to number of genome genes that are annotated with terms from the same database. GO: Gene Ontology. Supplementary Figure 6: Over-represented GO, DO and KEGG terms associated with CD38 in miR-140-3p-mediated regulatory network.

Dot plot for all the most interesting GO, DO and KEGG terms whose enrichment is determined by CD38. For an in-depth description of the plot see Supplementary Figure legend. For more information about what symbol shapes stand for see figure legend. GO terms whose enrichment is determined by both CD38 and NRIP1 are represented by bigger symbols. GO: Gene Ontology; DO: Disease Ontology; KEGG: Kyoto Encyclopedia of Genes and Genomes. Given its prevalence and social impact, Autism Spectrum Disorder (ASD) is drawing much interest. Molecular basis of ASD is heterogeneous and only partially known.

Many factors, including disorders comorbid with ASD, like TS (Tourette Syndrome), complicate ASD behavior-based diagnosis and make it vulnerable to bias. To further investigate ASD etiology and to identify potential biomarkers to support its precise diagnosis, we used TaqMan Low Density Array technology to profile serum miRNAs from ASD, TS, and TS+ASD patients, and unaffected controls (NCs). Through validation assays in 30 ASD, 24 TS, and 25 TS+ASD patients and 25 NCs, we demonstrated that miR-140-3p is upregulated in ASD vs.: NC, TS, and TS+ASD (Tukey's test, p-values = 0.03, = 0.01. IntroductionAutism Spectrum Disorder (ASD) is the name for a heterogeneous group of complex neurodevelopmental conditions, which are clinically defined by: (1) defects in social interaction and communication; (2) fixed interests and repetitive behaviors. Typically, ASD symptoms become fully manifest during school age and have a lifelong impact on everyday functions (American Psychiatric Association, ).The broadening of the autism concept and the resulting changes in ASD categorization have increased ASD awareness and improved its diagnostic surveillance in health and educational care.

This has led to an alarming rise in the number of milder cases of ASD, without co-occurring intellectual disability, in developed countries around the world.Recently, it has been reported that ASD affects one in 68 US children and that approximately four males suffer from ASD for every female (Christensen et al., ). Comorbid neuropsychiatric and neurodevelopmental disorders contribute to ASD impairment, being common (70.8%) and frequently multiple (57%) in ASD children (Simonoff et al., ). Such conditions include social anxiety disorder, attention-deficit/hyperactivity disorder (ADHD), oppositional defiant disorder (ODD), chronic tic disorder, and obsessive-compulsive disorder (OCD).Tourette Syndrome (TS) is a neurodevelopmental disorder characterized by considerable motor as well as behavioral impairment: it affects approximately 1% of the population with a male:female ratio of 3:1.

It is clinically defined by childhood onset of multiple motor tics and at least one phonic tic, which collectively must persist for at least 12 months (American Psychiatric Association, ). 88% of TS patients also show comorbidity and psychopathology. Comorbidity with ADHD and OCD is very common.

Co-existent psychopathologies include depression, anxiety, learning difficulties, personality disorder, ODD, and conduct disorder (Robertson, ). TS can cause severe difficulties in social functioning and a reduced quality of life in patients suffering from it. Characteristic, but not essential for diagnosis, symptoms include complex tics, such as echolalia and echopraxia (copying others' vocalizations and actions, respectively), palilalia and palipraxia (repeating own words/phrases and actions, respectively), coprolalia (inappropriate involuntary swearing) as well as repeating of complex words (Robertson, ).It has been observed that 4.8% of ASD children also suffer from TS (Simonoff et al., ) and that 6–11% of TS cases show comorbidity with ASD (Robertson, ). TS and ASD share clinical symptomatology and many behavioral features. Genetic studies also support the existence of common susceptibility genes in both disorders (Clarke et al., ). The exact etiology of both disorders is still elusive.Strong evidence suggests that ASD may arise from genetic, epigenetic and environmental factors (Abdolmaleky et al.,; Nardone and Elliott,; Sun et al., ). ASD is genetically highly heterogeneous.

Both inherited and de novo ASD-associated variants have been characterized in hundreds of genes. Both inherited and de novo rare genetic variants can be detected in 10–30% of ASD cases. Single common inherited variants can be found in approximately 1.1–1.2% of ASD cases; when considered cumulatively, these can explain 15–50% ASD cases. Patient selectionFrom a database of more than 2,000 patients (from the Section of Child and Adolescent Psychiatry, Department of Clinical and Experimental Medicine, University of Catania), 79 Caucasian patients, aged 3–13 years from various socio-economic contexts, were randomly recruited and studied from January to November 2016. Thirty patients affected by ASD mean age 6.5 (standard deviation, SD 3.5); M:F 22:8, 24 patients affected by TS mean age 8.7 (SD 5.2); M:F 21:3, and 25 patients affected by TS+ASD (mean age 9.3 (SD 6.7); M:F 25:0 were included in the study.

They were compared to 25 neurologically intact unaffected negative control (NC) individuals mean age 9.5 (SD 3.9); M:F 16:9, recruited from local schools, without any history of either ASD or TS and who suffered from neither chronic diseases nor psychiatric disorders (Table ). Data are shown as means and standard deviations between parentheses. ASD, Autism Spectrum Disorder; TS, Tourette Syndrome; NC, Unaffected Control; M, male; F, female; IQ, Intelligence Quotient; YGTSS, Yale Global Tic Severity Scale; ADOS, Autism Diagnostic Observation Schedule.The study was approved by the local Ethics Committee.

All parents gave written informed consent.Diagnoses of ASD, TS and other clinical conditions were made according to both DSM-IV-TR (Diagnostic and Statistical Manual of Mental Disorders, IV edition–Text Revision) and DSM-5 criteria by a child neurologist (RR). All the participants were evaluated at the University Hospital Policlinico - Vittorio Emanuele of Catania. The three clinical groups (ASD, TS, and TS+ASD) and the NCs were assessed using the following scales/schedules: ADOS and ADI-R to evaluate ASD symptoms; YGTSS to evaluate presence and severity of tics. Moreover, the three clinical groups (ASD, TS, and TS+ASD) and the NCs were also assessed by a psychologist through WISC-III (Wechsler Intelligence Scale for Children, III edition) as an evaluation of both IQ (Intelligence Quotient) and cognitive functioning. Neuropsychological features of participants are summarized in Table.In a previous study (Rizzo et al., ), we reported that only miR-429 was significantly differentially expressed (DE) in the serum of TS patients compared to NCs: TS patients were included in this experimental series aiming to compare them with the other classes of neuropsychiatric patients. Sample processingPeripheral blood samples from all participants were taken in the morning using a butterfly device into serum separator collection tubes, provided with Clot activator and gel for serum separation as additives (BD Biosciences).

Collection tubes were treated according to current procedures for clinical samples. In order to separate serum from blood cells, tubes were rotated end-over-end at 20°C for 30′ and then centrifuged at 3,500 rpm at 4°C for 15′ in a Beckman J2-21.

Supernatants were aliquoted into 1.5 ml RNase-free tubes and stored at −80°C. Prior to RNA extraction, stored supernatants were centrifuged again at 3500 rpm at 4°C for 15′ to remove circulating cells or debris. Serum samples were aliquoted into 1.5 ml RNase-free tubes and they were either immediately used for RNA extraction or stored at −80°C until analysis (Rizzo et al., ).

MiRNA profilingWe used TLDA (TaqMan Low Density Array) technology to profile the serum expression of 754 different human miRNAs of four ASD patients, five TS patients, four TS+ASD patients and three NCs. 3 μl of RNA were reverse transcribed and preamplified according to manufacturer's instructions. Preamplified products were loaded on TaqMan Human MicroRNA Array v3.0 A and B 384-well microfluidic cards (Applied Biosystems, Foster City, CA, USA). PCR reactions on TLDAs were performed on a 7900HT Fast Real Time PCR System (Applied Biosystems) (Ragusa et al., ).We individually carried out the analysis on microfluidic cards A and B. We used a customized normalization approach for the relative quantification analysis. Supplementary File reports Ct values for TLDA panels A and B and shows our procedure step by step. For each comparison, a Ct value matrix (miRNAs in rows, samples in columns) was created.

In a similar way to the GMN (global median normalization) method (Park et al., ), for each sample of the comparison, the median and mean Ct values within the array, reflecting the loaded mass of template cDNA, were calculated. However, all Ct values representing a specific miRNA were kept out of these calculations if even just one of them corresponded to a flagged well. Then, using the Pearson correlation, miRNAs whose expression profile was closer (more positively correlated) to these values were identified as the best endogenous controls within the arrays. We normalized MiRNAs to the top three stable miRNAs within the arrays. MiR-146a and miR-223.

were the most frequently stable miRNAs for cards A and B, respectively, and the most abundant among those we could select.Therefore, for each comparison, three ΔCt value matrices (miRNAs in rows, samples in columns) were produced according to the 2 −ΔΔCt method (Schmittgen and Livak, ). DE circulating miRNAs were obtained performing SAM (Significance of Microarrays Analysis) statistical analysis on these matrices with MeV (Multi experiment viewer v4.8.1) statistical analysis software. For each pairwise comparison, we used a two-class unpaired test, based on at least 100 permutations per miRNA, with a FDR (False Discovery Rate) cut-off of 0.15, in order to detect dysregulated miRNAs. This analysis identified many DE miRNAs for each comparison. However, we have used very strict criteria to select miRNAs for further validation (i.e., number of SAM tests in which they were identified as DE, number of comparisons in which they resulted as DE, their abundance and the quality of their amplification curves during the profiling runs) in order to investigate only the most promising ones.

MiRNA profiling data validationRNA from sera of 30 ASD, 24 TS, and 25 TS+ASD patients and 25 NCs was used to perform miRNA-specific reverse transcription reactions producing miRNA-specific cDNAs for real-time PCRs. These RT-PCR analyses were performed using TaqMan MicroRNA Assays (Applied Biosystems) specific for the most interesting miRNAs identified as DE, miR-30d, miR-140-3p, miR-148a., and miR-222, and for the selected endogenous control, miR-146a. At first, the ASD group was composed of 32 patients. We checked if some of those samples should be considered as outliers, within this original ASD group, for: (1) the serum expression of miR-140-3p; (2) the severity of ASD symptoms.

We have looked at their ΔCt values for miR-140-3p and at scores that they obtained for the four items of the ADOS scale (A: Communication; B: Social interaction; C: Imagination; D: Repetitive and restricted behaviors). For these expression values and ADOS scores, we defined the corresponding mean ± 2. (SD) ranges and we considered patients with a value and/or score outside of those ranges as outliers. Two ASD patients were excluded from the original ASD group since: (1) both were outliers for miR-140-3p expression; (2) one was an outlier for the Imagination item (1/4 ADOS items), whereas the other one was an outlier for the Imagination and Repetitive and restricted behaviors items (2/4 ADOS items). All the following analyses were performed with GraphPad Prism for Windows v6.01 (GraphPad Software, La Jolla California USA ).

D'Agostino-Pearson omnibus K2 test and Shapiro-Wilk test were performed to check if data from every small group were normally distributed. Ordinary one-way ANOVA was used to test the differential expression of the selected miRNAs between the four groups. Statistical significance was established at a p ≤ 0.05. Tukey's multiple comparisons test was performed to identify which groups differed in the selected miRNAs' expression. Statistical significance was established at a multiplicity adjusted p ≤ 0.05.

Expression FC (Fold Change) values of miRNAs were calculated by applying the 2 −ΔΔCt method (Schmittgen and Livak, ). Correlation between miR-140-3p expression and neuropsychiatric parametersCorrelation between ΔCt values for miR-140-3p, obtained from the normalization to miR-146a, and neuropsychiatric parameters was analyzed in both a general (all patients and controls) and a class-specific (just one class of patients and controls) way, since some of these parameters were related to a certain class of neuropsychiatric disorders. IQ, ADOS items regarding communication, social interaction, imagination, and repetitive and restricted behaviors (ADOS items A-D), and YGTSS (Yale Global Tic Severity Scale) were the neuropsychiatric parameters chosen for this analysis. Either Pearson or Spearman correlation was computed on GraphPad Prism software when analyzing normally and not normally distributed data, respectively. Two-sided p-values from this correlation analysis were corrected for multiple comparisons by using three different approaches: Bonferroni correction, Holm-Bonferroni correction, and Benjamini-Hochberg (BH) FDR procedure. Statistical significance was established at a p-value ≤ Bonferroni corrected α = 0.05/16 = 0.003, at a Holm-Bonferroni corrected p ≤ 0.05, and at a Benjamini-Hochberg FDR adjusted p ≤ 0.01. Linear regression analysis was also carried out on GraphPad Prism software only for significant correlations.

Statistical significance was established at a p ≤ 0.05. Reconstruction of the miR-140-3p-mediated regulatory networkMiR-140-3p targets whose validation was based on strong evidence were retrieved by DIANA-TarBase v7.0 (Vlachos et al., ) and miRTarBase (Chou et al., ) databases. The biological network, composed of MIR140, these targets, and their first neighbors, was built retrieving interactome data through BisoGenet v3.0.0 Plug-in (Martin et al., ) in Cytoscape v3.4.0 (Shannon et al., ).

Network centralities analysis, permitting the identification of the nodes that, more than others, were good candidates as regulators of the underlying biological processes in which the network is involved, was carried out through CentiScaPe v2.1 Plug-in (Scardoni et al., ). Network functional analysisclusterProfiler v3.2.11 R package (Yu et al., ) was used to perform functional enrichment analyses on miR-140-3p-mediated regulatory network node genes in R v3.3.2 (R Core Team, ). We searched for the gene annotation terms from the GO (Gene Ontology), DO (Disease Ontology), KEGG (Kyoto Encyclopedia of Genes and Genomes), and Reactome databases that were over-represented in the list of network node genes compared to the entire genome. Statistical significance for the hypergeometric test was established at a BH adjusted p ≤ 0.05.

Gofilter and simplify functions in clusterProfiler were employed in order to select level-specific GO terms and to remove the most redundant ones, respectively. Network expression analysisIn order to investigate if deregulation of network node genes was implicated in ASD, we searched for raw high-throughput gene expression datasets produced from the analysis of samples of ASD patients on two public repositories, GEO (Gene Expression Omnibus) DataSets (Edgar et al., ) and ArrayExpress (Kolesnikov et al., ). Datasets retrieved by GEO DataSets were analyzed performing limma tests with the GEO2R tool (Barrett et al., ), whereas datasets retrieved by ArrayExpress were analyzed performing Tusher SAM tests with MeV software. We reported only network node genes whose log 2FC expression was significantly higher than 1 and lower than -1 as upregulated and downregulated, respectively, within ASD datasets. MeV software was also used to produce the curated ASD expression heatmap. Supplementary File reports all the datasets selected for network functional analysis along with their references and IDs (Gregg et al.,; Kuwano et al.,; Voineagu et al.,; Ginsberg et al.,; Kong et al.,; see Supplementary File ). ROC curve analysis and biomarker performance evaluationΔCt values for miR-140-3p, obtained from the normalization to miR-146a, served as input data to perform a classical univariate ROC (Receiver Operator Characteristic) curve analysis for each of the comparisons where we found this miRNA to be dysregulated on the server Metaboanalyst 3.0 (Xia and Wishart, ).

An appropriate ΔCt cut-off point maximizing both sensitivity and specificity (that is, the threshold that maximizes the distance to the diagonal line) was found for each curve by calculating the maximum Youden index J (max (sensitivity + specificity)–1). GraphPad Prism software was used to create Figure 4. The true positive rate (y-axis) was plotted in function of the false positive rate (x-axis), for different ΔCt cut-off points.Since these ROC curves were based on a miRNA already identified as differentially expressed between the compared groups (miR-140-3p), through them we could only assess its idealized discriminative power. It is possible that this miRNA only accurately predicts outcomes in the initial data set and that minor fluctuations in the training data could markedly lower its predictive performance.Therefore, after these preliminary ROC curve analyses, we built corresponding logistic regression models for the expression of miR-140-3p and we tested them through CV (cross-validation) and permutation testing, once again, by using the server Metaboanalyst 3.0. CV gives an indication of how accurate a given model might be in predicting new samples, validating its general applicability (Xia et al., ).

100-time repeated random sub-sampling CV was used to test the performance of the built logistic regression models. At each CV, 2/3 of samples are used for model training and the remaining 1/3 of samples are used for testing. Permutation testing indicates if a given model is significantly different from a random guessing model for the sample population, validating the proposed model structure (Xia et al., ). Permutation testing on the performance measure AUC (Area under the ROC curve) was used to calculate the significance of the built logistic regression models. The permutation tests use this procedure: random label re-assignment to each sample; 3-time repeated random sub-sampling CV; comparison of the performance measures between the models obtained by using the original and the permuted sample labels. This procedure was repeated 100 times. If the performance measure of the original data lies outside the normal distribution of the one of the permuted data, then the tested model is significant.

Statistical significance was established at a p ≤ 0.05. High-throughput expression analysis of circulating miRNAs in ASD, TS, and TS+ASD patientsBy using TLDA technology, we investigated the expression levels of 754 miRNAs in sera from four ASD patients, five TS patients, four TS+ASD patients, and three unaffected NCs. We identified miR-146a and miR-223. as the best endogenous controls for panels A and B, respectively.

Supplementary File reports Ct values for TLDA panels A and B.We found that four miRNAs from panel A (miR-140-3p, miR-222, miR-454, miR-483-5p), and five miRNAs from panel B (miR-30d, miR-30e-3p, miR-148a., miR-1274B, miR-1290) were significantly DE in at least one of the comparisons made (FDR. Dysregulated expression levels of miR-140-3p in serum from ASD patientsWe selected miR-30d, miR-140-3p, miR-148a., and miR-222 for further validation through single TaqMan assays.

MiR-146a was used as endogenous control in all the analyses carried out. Supplementary File reports all ΔCt values from validation assays.We found only miR-140-3p as significantly dysregulated in serum from ASD patients (ordinary one-way ANOVA, p = 0.0001). In particular, serum levels of miR-140-3p were higher in 30 ASD patients compared to 25 NCs (Tukey's multiple comparisons test, multiplicity adjusted p = 0.03), 24 TS patients (Tukey's multiple comparisons test, multiplicity adjusted p = 0.01), and 25 TS+ASD patients (Tukey's multiple comparisons test, multiplicity adjusted p.

ALL PATIENTS AND NCSΔCt vs. ADOS CommunicationΔCt vs. ADOS Social interactionΔCt vs.

ADOS ImaginationΔCt vs. ADOS Repetitive and restricted behaviorsSpearman r0.020.33−0.13−0.07−0.17NTPearson rNTNTNTNTNT−0.1795% CI−0.18–0.210.15–0.50−0.32–0.07−0.27–0.12−0.36–0.03−0.35–0.02two-sided p-value0.860.00050.180.450.090.08Is p. We performed the analyses in both a general (Section 1) and a class-specific (Sections 2–4) manner. Either Pearson or Spearman r values from every analysis are reported. Bonferroni corrected α = 0.05/16 = 0.003.

Correlation between serum levels of miR-140-3p and scores from YGTSS scale. The scatterplot refers to all the 104 analyzed samples and it also reports the best-fit line obtained from linear regression analysis. YGTSS, Yale Global Tic Severity Scale; 95% CI, 95% Confidence Interval.This analysis confirmed that serum expression of miR-140-3p correlated with a crucial neuropsychiatric scale for the clinical diagnosis of TS. We infer that miR-140-3p could prove to be useful to strengthen the behavior-based diagnosis of either ASD or TS+ASD, which can be particularly challenging in some clinical cases.

Reconstruction of miR-140-3p-mediated regulatory network: functional and expression analyses of network node genesBy searching on online databases of miRNA-mRNA interactions for validated targets of miR-140-3p, we retrieved CD38 ( CD38 molecule) and NRIP1 ( nuclear receptor interacting protein 1) as its only targets whose validation was based on strong evidence. Through network analysis, we reconstructed the regulatory network composed of MIR140 ( microRNA 140, the gene encoding miR-140-3p), CD38, NRIP1, and their first neighbors. This network had 111 nodes and 821 edges. NRIP1, POLR2A ( RNA polymerase II subunit A), EP300 ( E1A binding protein p300), E2F1 ( E2F transcription factor 1), ESR1 ( estrogen receptor 1), PHF8 ( PHD finger protein 8), and TAF1 ( TATA-box binding protein associated factor 1) were the nodes with the highest degree within it (Supplementary Figure ).In order to investigate the potential etiological role of this miRNA-mediated network in ASD, we performed functional enrichment analysis of network node genes using GO, DO, KEGG, and Reactome gene annotation databases (Supplementary Figures –). Supplementary File reports all the data from network functional analysis.Genes from miR-140-3p-mediated regulatory network played a role in various mechanisms within the nervous system (i.e., neurogenesis, regulation of synaptic plasticity, long term synaptic depression, cellular response to nerve growth factor, neuron differentiation, dendrite development, and neuronal death). In addition to their role in nervous system development, they were also involved in growth regulation, endocrine system development, heart development, respiratory system development, and tongue development (Table, Supplementary Figure ).

MIR140 gene was annotated with the over-represented DO term physical disorder, that refers to diseases determined by a genetic abnormality, error with embryonic development, infection or compromised intrauterine environment (DOID:0080015; BH adjusted p = 0.027; Gene Ratio = 0.070; Background Ratio = 0.017). Among the most interesting terms whose enrichment was determined by CD38, we found those regarding response to estradiol, retinoic acid, drugs, hypoxia, ketone, and oxidative stress, activation and proliferation of immune cells, regulation of protein localization, cellular calcium ion homeostasis, and blood circulation.

We found bacterial infectious disease as the only DO term among them. Finally, CD38 was directly involved in the regulation of synaptic plasticity after and in long-term synaptic depression (Table, Supplementary Figure ). Among the most interesting terms whose enrichment was determined by NRIP1, there were those related to response to estradiol and steroid hormones, reproductive system development, and development of primary sexual characteristics. NRIP1 was annotated with many molecular functions (i.e., histone deacetylase binding, nuclear hormone receptor binding, core promoter sequence-specific DNA binding, retinoic acid receptor binding, and retinoid X receptor binding).

Setup torrent tracker site. Finally, NRIP1 regulated the transcription of genes involved in circadian rhythms by interacting with RORA ( RAR related orphan receptor A) (Table, Supplementary Figure ). Term database ID, term description, corresponding BH adjusted p-value generated by the hypergeometric test, Gene Ratio, and Background Ratio values are reported for all the GO Biological Process terms regarding nervous system and development. Gene Ratio: the ratio of number of genes of interest that are annotated with a certain term from the database used to perform the analysis to number of genes of interest that are annotated with terms from the same database. Background Ratio: the ratio of number of genes in the genome that are annotated with a certain term from the database used to perform the analysis to number of genome genes that are annotated with terms from the same database.

GO, Gene Ontology; BP, Biological Process; BH, Benjamini-Hochberg. Term database, term ID, term description, corresponding BH adjusted p-value generated by the hypergeometric test, Gene Ratio, and Background Ratio values are reported for all the over-represented terms with which NRIP1 is annotated.

For an in-depth description of the column names see Table legend. GO, Gene Ontology; BP, Biological Process; MF, Molecular Function; CC, Cellular Component; DO, Disease Ontology; BH, Benjamini-Hochberg.To verify if dysregulation of network node genes was implicated in ASD, we used publicly available raw high-throughput gene expression datasets produced from the analysis of ASD samples. Ten genes were found to be downregulated in whole blood of ASD patients compared to NCs (log 2FC 1): PHF8 and TAF1 (Figure ). Supplementary File reports all the data from network expression analysis. Expression analysis of node genes within miR-140-3p-mediated regulatory network in human ASD high-throughput gene expression datasets retrieved from GEO DataSets and ArrayExpress. Datasets used for the expression analysis (human ASD source tissue, dataset ID, platform type, statistical test performed) and microarray probe IDs along with their corresponding gene symbols are reported in columns and lines of this gene expression heatmap, respectively.

Colored heatmap cells represent genes that are DE in a certain dataset. Data are shown as log 2FC expression values.

For more information about gene expression trend, see figure legend. Serum levels of miR-140-3p in the discrimination of ASD patientsWe used ΔCt values for miR-140-3p to perform a classical univariate ROC curve analysis for each of the comparisons where we found this miRNA to be dysregulated. The univariate ROC plots revealed an AUC of 0.71 for the comparison ASD vs. NC ( p = 0.006), 0.73 for ASD vs. TS ( p = 0.002), and 0.78 for ASD vs. TS+ASD ( p = 0.00007) (Figure ). We used ΔCt value cut-offs corresponding to the sensitivity/specificity pair with the highest Youden index J for every computed ROC curve to perform a blind diagnosis on all the 104 analyzed samples (Figure ).

Classical univariate ROC curve analyses for the comparisons in which miR-140-3p is dysregulated. This graph compares three ROC curves, one for each comparison where we found miR-140-3p to be dysregulated. Each point on the ROC curves represents a sensitivity/specificity pair corresponding to a particular decision threshold (ΔCt value cut-off).

Circles on the curves refer to the sensitivity/specificity pairs with the highest Youden index J. AUC, Area under the ROC curve; 95% CI, 95% Confidence Interval. The potential use of serum miR-140-3p as a biomarker: criteria for ASD diagnosis. The graphs show the distribution of ΔCt values of all the 104 analyzed samples, for which we already had a clinical diagnosis. We used data from classical univariate ROC curve analyses to perform a blind diagnosis of all study participants. In (A), the ΔCt ≤ 2.427 criterion divides ASD patients from NCs and determines the correct discrimination of 19/32 ASD patients and 20/25 NCs.

In (B), the ΔCt ≤ 2.447 criterion divides ASD patients from TS patients and determines the correct discrimination of 19/32 ASD patients and 19/24 TS patients. In (C), the ΔCt ≤ 2.824 criterion separates ASD patients from TS+ASD patients and determines the correct discrimination of 22/32 ASD patients and 19/25 TS+ASD patients.Then, we built a logistic regression model for miR-140-3p expression in each comparison and we tested those predictive models through CV and permutation testing. 100-time repeated random sub-sampling CV was used to test the performance of the logistic regression models. MiR-140-3p continued to perform at a good level for the comparison ASD vs. NC, with an average AUC of 0.70, a sensitivity of 63.33%, and a specificity of 68% (Figures ).

MiR-140-3p continued to perform at a good level also for the comparison ASD vs. TS, with an average AUC of 0.72, a sensitivity of 66.66%, and a specificity of 70.83% (Figures ). MiR-140-3p continued to perform at a very high level for the comparison ASD vs. TS+ASD, with an average AUC of 0.78, a sensitivity of 73.33%, and a specificity of 76% (Figures ). CV results demonstrated the general applicability of these predictive models. 100-time repeated permutation tests on the performance measure AUC were carried out to validate the structure of these models. Permutation testing results were significant and quite stable in different runs for all the models tested (Figures ).

Serum miR-140-3p could be used to discriminate ASD patients. (1) The graphs refer to the comparison ASD vs.

(2) The graphs refer to the comparison ASD vs. (3) The graphs refer to the comparison ASD vs. (A) Classical univariate ROC curve analysis. The red dot represents the sensitivity/specificity pair with the highest Youden index J. (B) Boxplot depicting the distribution of ΔCt values in the two groups. The red line represents the ΔCt value cut-off corresponding to the red dot on the curve in (A).

The label 1 refers to the ASD group, 0 to the other group. (C) Average ROC curve from 100-time repeated random sub-sampling CV of the built logistic regression model. (D) Average predicted class probabilities (x-axis) of each sample (y-axis) from the 100 CV iterations. Probability scores more than 0.5 belong to the ASD group, those less than 0.5 belong to the other group. Incorrectly classified subjects are identified by their ID number.

(E) Results from the permutation tests on the model performance measure AUC. Average ROC curve and corresponding p-value are reported. AUC, Area under the ROC curve; 95% CI: 95% Confidence Interval; CV, cross-validation.These data proved that serum miR-140-3p could be used in the discrimination of ASD patients. In particular, it could potentially support the differential behavior-based diagnostic process of two classes of neurodevelopmental disorders, ASD and TS+ASD. Supplementary Figure 2Over-represented GO terms in miR-140-3p-mediated regulatory network compared to the entire genome. (A) Dot plot for GO Biological Process terms.

(B) Dot plot for GO Molecular Function terms. (C) Dot plot for GO Cellular Component terms. Each circle in the plots symbolizes an over-represented term: its x-axis coordinate reflects the Gene Ratio value; its size is directly proportional to the Count value; its color represents the Benjamini-Hochberg adjusted p-value generated by the hypergeometric test.

Gene Ratio: the ratio of number of genes of interest that are annotated with a certain term from the database used to perform the analysis to number of genes of interest that are annotated with terms from the same database. Count: number of node genes within the network that are annotated with a certain term. GO: Gene Ontology. Supplementary Figure 5Over-represented GO Biological Process terms regarding nervous system and development in miR-140-3p-mediated regulatory network. Dot plot for all the most interesting GO Biological Process terms regarding nervous system and development.

Enciclopedia Taurina Cossio Pdf English

Each symbol in the plot symbolizes an over-represented term: its x-axis coordinate reflects the Gene Ratio value; its y-axis coordinate reflects the Background Ratio value; its color represents the Benjamini-Hochberg adjusted p-value generated by the hypergeometric test. Gene Ratio: the ratio of number of genes of interest that are annotated with a certain term from the database used to perform the analysis to number of genes of interest that are annotated with terms from the same database. Background Ratio: the ratio of number of genes in the genome that are annotated with a certain term from the database used to perform the analysis to number of genome genes that are annotated with terms from the same database.

GO: Gene Ontology. Supplementary Figure 6Over-represented GO, DO and KEGG terms associated with CD38 in miR-140-3p-mediated regulatory network.

Dot plot for all the most interesting GO, DO and KEGG terms whose enrichment is determined by CD38. For an in-depth description of the plot see Supplementary Figure legend. For more information about what symbol shapes stand for see figure legend. GO terms whose enrichment is determined by both CD38 and NRIP1 are represented by bigger symbols. GO: Gene Ontology; DO: Disease Ontology; KEGG: Kyoto Encyclopedia of Genes and Genomes.