Skip to main content
Fig. 2 | BMC Bioinformatics

Fig. 2

From: CANTARE: finding and visualizing network-based multi-omic predictive models

Fig. 2

Cumulative fit for Model S. This model was created from the Subdoligranulum hub. It has an AUC of 0.93 and the fewest predictors (7) of the top 3 models. The cumulative fit graph displays predicted probability by accumulating the contribution of one predictor at a time. A dotted black “trend line” connects the mean cumulative value at each predictor. Omic predictors are prefaced by met or mb, representing metabolite or microbe respectively. Each orange path represents one study participant with IBD, while each green path represents one control participant. The intersection of the paths and the bold black vertical line represents predicted probability for each person, while the final points of the paths indicate the actual class. The predictors are ordered by effect size, low to high. The intercept is not shown. An isomer of urobilin is indicated by “urobilin*”. Effect sizes for each predictor are expressed as the odds ratio for IBD for either interquartile range (IQR) or overall range. If the first quartile equals the third quartile, the odds ratio is expressed in terms of overall range. The IQR effect size is the change in response associated with a change in a predictor from the first quartile (Q1, 25th percentile) to the third quartile (Q3, 75th percentile), which includes 50% of the data values

Back to article page