Fig. 3. LICTOR accurately predicts the LC toxicity of sequences absent from the training set and is able to revert the pathological phenotype of a cardiotoxic LC.
a LICTOR predictions based on an independent set of LCs, i.e. not present in the training set. Toxic LCs are from patients affected by AL with cardiac involvement, while non-toxic LCs are from patients with multiple myeloma (see also Supplementary Data 2). Predictions are divided according to the clinical phenotype. White part of the bars represent correct LICTOR predictions, while grey part of the bars represent incorrect predictions. b Sequence of a cardiotoxic LC (tox153) used to neutralize the toxic phenotype using LICTOR and the non-toxic features unveiled by feature selection. Tox153 is aligned with the corresponding germline (GL) sequence, and the third line shows the difference in somatic mutations (SMs) between the LC and the GL sequence (LC/GL) with SMs (displayed in bold) and unmutated positions represented by an “X”. c The table represents the non-toxic features according to information gain used to revert the toxic phenotype of tox153. For each predictor variable, we also report the ranking in the specific-feature family (a) and the feature selection general ranking (b). d Proteotoxic effect of tox153 protein, of the two mutants in silico designed by adding non-toxic features (tox153V52L and tox153V52LA56G) and of the tox153 GL protein. The proteotoxic effect of H18 cardiotoxic LC and of the GL proteins H6GL and H9GL are tested as well. Proteins in 10 mM PBS (100 μg/mL) were administered to C. elegans (100 μL/100 worms). Vehicle (10 mM PBS) and 1 mM H2O2 were administered as negative and positive controls, respectively. Pharyngeal activity was determined 24 h after treatment by determining the number of pharyngeal bulb contractions (pumps/min). Data are the pumps/min in mean ± SE (n = 30 worms/assay, two assays). ****p < 0.0001 one-way ANOVA, Dunn’s post hoc test. e Values of pharyngeal bulb contraction (pumps/min) of some LCs (H6, H7, H9 and M2, M7, M8) listed in Supplementary Data 2 are from AL patients with cardiac involvement (cardiotoxic) and patients with multiple myeloma (non-toxic). These values were previously obtained under the same experimental conditions employed in this study35,36. Additionally, values of pharyngeal bulb contraction (pumps/min) of H18, H6GL and H9GL are reported as well (n = 30 worms/assay, two assays). Each square is the mean value for H6, H7, H9 and H18 while dots represent the mean of M2, M7, M8, H6GL and H9GL. Horizontal lines represent the mean of cardiotoxic and non-toxic LCs (****p < 0.0001, two-tailed unpaired t-test). The values of pumps/min obtained after the administration of tox153, tox153V52L and tox153V52LA56G are also plotted (triangles).