Two samples (B475 and B22) were highly active, as active as the standard haemorrhagic

venom (B. jararaca). Venom samples from B208, B33, B67 and B5 were moderately active (compared to B. jararaca), while those from B8, B469 and A229 were of low haemorrhagic activity. Myotoxic activity was rare and usually mild. Only T224, T221 and T61 (fraction 20) were clearly myotoxic although B526 and T208 were mildly myotoxic. Oedema was common, but non-specific ( Table 1). Clear evidence of neurotoxicity was seen only with T61 (fraction 20) ( Table 1, Fig. S1). The total dataset contained 253 non-redundant protein sequences (Fig. 1). The alignment is available in the Dryad data depository (doi:10.5061/dryad.16pg7). The click here first four factors describing amino acid composition were retained. These PS-341 purchase principal components, referred to as PC1-4(comp) hereafter, summarised 16.7, 14.0, 10.1, and 9.5% of variation respectively. Ninety-five proteins had known functions that could be assigned to one of six major functions. However, anticoagulant and antiplatelet functions were subsequently combined into a “haemotoxic” category after preliminary analyses showed that no physico-chemical property or PC(comp) could distinguish between these groups (Tamhane’s post-hoc test). Final sample sizes were: Myotoxic: 30; Haemotoxic: 19; Neurotoxic: 26; Hypotensive: 7; Oedematous: 15. Neurotoxic PLA2s frequently also show myotoxicity

(Montecucco et al., 2008), but were classed as neurotoxic rather than myotoxic for the purpose of this analysis. Robust tests (Brown-Forsythe) for the equality of means showed all variables apart from PC2(comp) showed significant differences among groups. The four resulting discriminant functions (Table 2) contained 69.1, 13.5, 11.1, and 6.3% of variation respectively. Another 158 proteins which had no known function were plotted on the resulting axes (Fig. 2) and colour-coded by their posterior probabilities of belonging to one of the functional

groups (Table S2). All groups, except for haemotoxic and hypotensive proteins, were successfully discriminated on two axes (Fig. 2A). DF1 largely reflects the difference in pI, with haemotoxic/hypotensive proteins being acidic, myotoxic, neurotoxic and most oedematous proteins being basic. However, notably some oedematous BCKDHA proteins can be distinguished from myotoxic ones by being more neutrally charged at pH7. DF2 (Table 2, Fig. 2A) largely distinguishes a smaller group of oedematous proteins on the basis of PC3(comp), with oedematous toxins having lower amounts of phenylalanine, arginine and tyrosine, and higher amounts of methionine and valine. DF3 (not shown) is influenced by a contrast between net charge and pI, and further distinguishes myotoxins proteins from oedematous proteins and neurotoxins, with myotoxins displaying a lower net charge for a given pI than the other types.