The network was defined according to the criteria defined by (STRING v9.1, [267]): (1) conserved neighborhood (when genes occur in the same neighborhood in genomes), (2) cooccurrence (presence vs. absence of linked proteins across species), (3) co-expression (when genes are co-expressed in species), (4) fusion (a fusion event results in a hybrid gene formed from two formerly separated genes), (5) experimental interactions that result from experimental data, (6) base data examination (deduced from the sequence), and (7) data from published scientific papers. The probability of association between proteins is computed after weighing the different criteria [273].

Thirty-eight proteins define a first network with a .70 confidence link that corresponds to a high-level association [273]. The network is shown in Fig. 4. Seven groups emerged from a visual inspection of the network:

The SYNNAP1-PTCHD1 association was not considered as a possible group at this state for the reasons that will be presented later.

Does this apparent distinction between groups support the hypothesis of an association between ASD with a unique signaling pathway? We addressed the question characterizing each group by its GO properties quantified by an enrichment program. The items were pooled in larger categories covered by cellular components, molecular mechanisms, and biological processes. The categories served to compute the between-group correlations, and then a factor analysis helped to depict the structure generated by the correlations (Table 1).

Two factors were considered according to Kaiser criterion. The first factor is almost a general factor suggesting that the groups described above belong to a unique signaling pathway. The factor is characterized by the high loadings of groups 1, 3, 4, and 5 to a lesser extent. The loadings of the groups 6 and 7 are at the limit of significance (.20). The first factor classes the groups according to impact on the neurotransmission processes as indicated by the enrichment program. The groups that have the highest loadings have the highest scores in neurotransmission processes. A composite score corresponding to neurotransmission processes is shown in Fig. 5. A second factor accounting for a smaller part of variance was defined by the high loadings of groups 2, 5, and 4. The groups have the highest scores in the composite index (regulation and metabolic processes) compared to groups 1, 3, 6, and 7 (Fig. 5). Groups 4 and 5 have, however, significant loadings with the first factor, as we failed to obtain a simple factorial structure (loading on one factor only).

The result indicates that the signaling pathway is directed to two interrelated categories of processes. The first include the synaptic mechanisms or more generally those involved in the neurotransmission mechanisms. These mechanisms are grouped in different categories indicated in Fig. 6. They are associated with most of the dysfunction observed in the ASD-associated genes (Sect. 3.1.2 impact of ASD-associated genes on neuron functions).

The second include the metabolic or regulatory mechanisms of the disorders which were reported as characterizing several of the ASD-associated genes.

Thirteen other proteins can be added to the previous list when the confidence link is relaxed (.40). We did not find any association between the proteins of the new list, but we found that they are associated with the proteins of the first list (Fig. 6). They cannot be considered therefore as modifying the bi-factorial structure found for the first 38 proteins. The SYNGAP1-PTCHD1 pair that was independent in the first group is associated now to the network via SHANK3. The last 12 proteins (ALDH7A1, BRAF, CNTN6, EHMT1, FOXP1, HOXA1, MBD5, NHS, NSD1, SATB2, SH3, and VPS13B) were neither interconnected nor associated with the proteins of the lists shown in Figs. 4 and 6. Their association with the present network or to new network could be generated by the discovery of new ASD-associated genes. In total, 51 out of the 63 genes associated with ASD form a unique signaling pathway.

The first and the second networks (Figs. 4 and 6) are based on observed or deduced association. They inform us on the meaning of vulnerability gene. A vulnerability gene or a susceptibility gene is a gene that increases the risk (probability) of the carrier to present the disease. A vulnerability gene does not determine the presence of the disease. Determining would mean that the probability to present the phenotype fits with genetic laws. We know that fragile X syndrome or Rett syndrome increases the risk of ASD. All the patients with Rett syndrome do not present ASD. Less than 18 % of patients with fragile X syndrome or 40 % of patients with Joubert syndrome (AHI1) present ASD. There is no determination at high probabilities to present the disorder. The difference between a carrier presenting the disease and the healthy carrier could depend on the allelic forms in the genetic background and particularly in the allelic forms carried by the ASD-associated genes. It should be revealing to observe the result of these interactions by analyzing the phenotype due to multiple-targeted ASD-associated genes.

It could be of interest to replace the information on association with information about interactions. Informing on interaction would open the road to a causal analysis of the network. Several species offer the possibility to analyze whether the proteins belonging to the network define a cascade. Mice, zebra fish, or Drosophila are relevant as long as the homology is observed. Yeast has been used to detect interaction networks between proteins in different disorders (see Chap. 1). It should be used consistently when a new ASD-associated gene is discovered.