Fig. 4: Consensus clustering integrated with machine learning modelling in AALS identifies subtype-specific target in ALS patients.

a Pipeline of potential target prioritization. PMA110 signature was interrogated against the differentially expressed proteins (DEPs) by comparing each subgroup to the control group, followed by supporting evidence of DEPs from ALS postmortem tissues. Images in Fig. 4a were created with BioRender.com. b Differentially expressed protein analysis (ALS Subtype vs Control) showing up- and down-regulated genes across all the four ALS subtypes. Proteins were considered as DEPs based on the threshold of fold change > 1.5 and adjusted p values < 0.0001. p values were calculated using a two-tailed t-test. Benjamini-Hochberg method was used for multiple testing correction to obtain the corrected p-value (padj). c Pie plots illustrating the percentage of subtype-specific DEPs and patient postmortem DEPs in PMA110 signature. d–f Boxplots showing the protein expression levels in proteomics subtypes of AALS dataset (top panels) and ALS postmortem tissue dataset (bottom panels). For box-and-whisker plot, the center line represents the median, the box bounds indicate the interquartile range (IQR) from the 25th to the 75th percentile, and the whiskers extend to 1.5 times the IQR below the first quartile and above the third quartile. The points beyond whiskers indicate outliers. p values were calculated using one-way ANOVA. For iPSC proteomic data, p values were adjusted for age at sampling using ANCOVA. *p < 0.05; ***p < 0.001; ****p < 0.0001; ns, not significant.