trations. For the screens we designed a comprehensive library targeting 254 human phosphatases. For each target gene, 3 different siRNAs were transfected. The full screen was performed 3 independent times using different passages of the 3 Phosphatases Modulating Neurite Outgrowth SH-SY5Y cells. The normalized activity values obtained for individual sequence-specific siRNAs were averaged over the screens, generating one activity-value per siRNA, which was used for subsequent statistical analysis. This approach was undertaken to reduce random error, thus increasing the sensitivity of the screen data. Data were normally distributed and calculated z-scores for all siRNAs are plotted in Analysis of Screen Data The widespread use of high-throughput siRNA-based screens has highlighted the unfortunate caveat of a high incidence of sequence-dependent OTEs, resulting in false positive hits. Phosphatases Modulating Neurite Outgrowth Conventional analysis methods are based on activity ranking followed by an arbitrary threshold cut-off selection process, which typically identifies hits with only very high activity. By contrast, redundant siRNA analysis is an alternative probability- based analysis approach, that takes into account the collective activity of all siRNAs targeting a specific gene, thus strongly reducing the chance of sequence-dependent OTEs. The method assigns a p-value to individual genes, which reflect the 5 Phosphatases Modulating Neurite Outgrowth distribution of all the gene-specific siRNAs tested toward high activity. This type of analysis thus identifies hits that are distributed much deeper into the MedChemExpress Aphrodine dataset than conventional activity cut-off methods while simultaneously yielding significantly better validation rates, which reflects a higher rate of identification of true biological hits in the primary analysis. In our dataset top 10 hits identified by RSA analysis also distributed much deeper into the dataset compared to top 10 activity hits. Validation 11 negative and 4 positive hits were selected for validation based on criteria of interest and representation of the various phosphatase subgroups. We predominantly selected negative modulators for validation and follow-up because their knockdown phenotype is more likely to be biologically specific, and in view of a potential interest as therapeutic targets. The validation screen was performed essentially as for the main screens, using siRNAs corresponding in sequence to one of the three siRNAs in the main screen, but purchased from another vendor. 9/11 of the negative regulators and 2/4 of the positive regulators tested were validated resulting in an overall validation efficiency of 73%. The knockdown efficiency of four randomly chosen hits, including both validated and non-validated hits was tested by qPCR. For 
all four genes specific siRNA treatment resulted in 5075% reduction in mRNA level, which indicates a reasonable knockdown efficiency, and thus that lack of phenotypic validation did not necessarily correlate with lack of mRNA knockdown. Even though we purposely designed the screen to enrich for hits affecting BDNF signaling we considered the possibility that some 26976569 hits would function in a BDNF- or TrkB-independent fashion. To distinguish between BDNF-dependent and -independent hits we performed further validation in absence of BDNF. Surprisingly, 26028783 we observed a significant effect of TrkB knockdown on neurite outgrowth in the absence of BDNF. This may reflect an autocrine