Nonetheless, closer examination shows that the contraindicative effect of glutamate (E), arginine (R) and lysine (K is not strong in the context of the algorithm and typical subsequences, so failure to predict is in practice more an effect of the subsequence as a whole). In the above and similar cases, substituting the lysine or arginine by, say, a serine (S) in “computer experiments” does not typically increase the score of any residue to more than 100. For example, in the case of the witchweed neuraminidase, changing RWR to SWS raised the peak value at the tryptophan to 99, a significant increase but still not exceeding 100. In some cases a value exceeding 100 can of course be attained. A false negative also found in later studies investigating these issues was the ox neuraminidase, with the subsequence DDHGVSWRYGGGVS containing the tryptophan (W) with peak value 96. Changing to an serine (S) the arginine (R) adjacent to the tryptophan (W) did have an effect, albeit that the only change was having the tryptophan as the only residue predicted, with a marginal score of 101.