Exploring the druggable space around the Fanconi anemia pathway using machine learning and mechanistic models

Abstract

Background: In spite of the abundance of genomic data, predictive models that describe phenotypes as a function of gene expression or mutations are difficult to obtain because they are affected by the curse of dimensionality, given the disbalance between samples and candidate genes. And this is especially dramatic in scenarios in which the availability of samples is difficult, such as the case of rare diseases.

Results: The application of multi-output regression machine learning methodologies to predict the potential effect of external proteins over the signaling circuits that trigger Fanconi anemia related cell functionalities, inferred with a mechanistic model, allowed us to detect over 20 potential therapeutic targets.

Conclusions: The use of artificial intelligence methods for the prediction of potentially causal relationships between proteins of interest and cell activities related with disease-related phenotypes opens promising avenues for the systematic search of new targets in rare diseases.

Keywords: Big data; Fanconi anemia; Genomics; Machine learning; Mathematical models; Signaling pathways.

Fig. 1

Fanconi anemia curated map, based in the KEGG FA pathway. There are two protein complexes: RPA, composed of RPA1, RPA2, RPA3 and RPA4, and Core, composed of FANCM, FANCG, FANCL, FAAP100, FANCA, FANCB, UBE2T, STRA13, FANCC, FAAP24, HES1, FANCE, FANCF, BLM, RMI1, RMI2 and TOP3A. At the end of the effector nodes, whose names are taken for the circuits, a description of the main functionalities triggered by the signaling circuits can be found

Fig. 2

Schema of the procedure followed for the analysis

Fig. 3

Observed distribution of circuit activities in the comparison between healthy and FA bone marrow cells

Fig. 4

Observed distribution of circuit activities in blood, a tissue affected by the disease, two tissues with a high rate of cell replication (skin and gastrointestinal), where DNA reparation is expected to play a relevant role and another tissue with low rate of cell replication (brain)

Fig. 5

Distributions of the cross-validation of the relevance values for the top 50 most relevant genes ordered by their mean. Above the relevance value of 0.006 the relevance rendered by the ML procedure and the means obtained from the cross-validation are consistent. Then this value is taken as a threshold

Fig. 6

the distribution of the R₂ score for each signaling circuit of the FA pathway across all the training/test splits. The R₂ score goes from -infinite to 1, where 0 represents a model that always predicts the mean for each task and a perfect model has a score of 1

Fig. 7

Enrichment analysis with GO terms and rare diseases