Skip to main content

Table 3 ProfhEX performances compared against already existing tools

From: ProfhEX: AI-based platform for small molecules liability profiling

Platform

No. targets

Type

Metric

RMSE

AUC

Sn*

Sp*

BA*

ProfhEX

46

R

0.71 (0.09)

0.91 (0.04)

0.87 (0.23)

0.97 (0.04)

0.92 (0.11)

Cortés-Ciriano et al. [76]

14

R

0.72 (0.07)

–

–

–

–

Yao et al. [77]

44

C

–

0.91 (0.04)

–

–

–

Mayr et al. [31]

29

C

–

0.84 (0.15)

–

–

–

Bosc et al. [35]*

27

C

–

–

0.75 (0.1)

0.93 (0.04)

0.85 (0.04)

  1. No. targets indicate the number of targets in common with ProfhEX. Type “R” or “C” indicates the type of learning task, either regression or classification, respectively. Metrics Sn, Sp and BA stand for Sensitivity, Specificity and Balanced Accuracy, respectively. The standard deviation is reported in brackets
  2. *Statistics come from a blind evaluation of the dataset provided by the authors
Back to article page

Journal of Cheminformatics

ISSN: 1758-2946

Contact us