Table 1 State-of-the-art sets of core-fitness essential genes considered to benchmark CoRe

Hart2014

State-of-the-art reference set of core-fitness essential genes

A set of 360 genes presented in [33] and used as a classification template by BAGEL: a supervised computational framework for quantifying gene essentiality significance in pooled library screens [10, 24].

Large collection of shRNA gene dependency profiles analysed with a linear algebra approach.

Hart2017

State-of-the-art reference set of core-fitness essential genes

A set of 684 genes introduced in [31].

BAGEL reanalysis of 17 genome-scale knockout screens in human cell lines performed with different libraries.

Behan2019

State-of-the-art reference set of core-fitness essential genes

A set of 553 genes presented in [11].

ADaM analysis of a large collection of gene dependency profiles from CRISPR-screens of 325 human cancer cell lines from different tissue-lineages/cancer-types (now part of the Project Score database [28]), using a manually curated version of the Hart2014 set (the curated Hart2014 CFGs), as training. This was obtained by excluding from the Hart2014 set 34 genes, such as for example KRAS and CHD4, predicted to be cancer drivers by the intOGen pipeline [34, 35]

Sharma2020

State-of-the-art reference set of core-fitness essential genes

A set of 519 genes presented in [30].

Logistic regression approach (part of the CEN-tools software), which uses the BAGEL essential/never-essential genes as training sets, respectively the Hart2017 set and a set of 927 never-essential genes [10, 24]. This approach was individually applied to the dependency profiles from Project Score [28] and from the Broad DepMap portal [](https://depmap.org) (Release 19Q2). The final predicted set was composed of genes predicted as CFGs in the two analyses, excluding those in the training set.

For the comparison with the unsupervised methods, this set was joined with the Hart2017 set (used in its training phase), rising up to 1182 genes.