# Predicting the Efficacy of Novel Synthetic Compounds in the Treatment of Osteosarcoma via Anti-Receptor Activator of Nuclear Factor-κB Ligand (RANKL)/Receptor Activator of Nuclear Factor-κB (RANK) Targets

## TL;DR

This paper explores how to predict the effectiveness of new synthetic drugs for treating osteosarcoma using computational models.

## Contribution

The study introduces a novel quantitative model using gene expression programming to predict drug efficacy in osteosarcoma.

## Key findings

- Gene expression programming outperformed heuristic methods in predicting drug efficacy with higher correlation coefficients.
- A non-linear model using five descriptors was developed for predicting IC50 values of synthetic compounds.
- The GEP algorithm showed better predictive accuracy for novel compounds compared to the HM algorithm.

## Abstract

Osteosarcoma (OS) currently demonstrates a rising incidence, ranking as the predominant primary malignant tumor in the adolescent demographic. Notwithstanding this trend, the pharmaceutical landscape lacks therapeutic agents that deliver satisfactory efficacy against OS.

This study aimed to authenticate the outcomes of prior research employing the HM and GEP algorithms, endeavoring to expedite the formulation of efficacious therapeutics for osteosarcoma.

A robust quantitative constitutive relationship model was engineered to prognosticate the IC50 values of innovative synthetic compounds, harnessing the power of gene expression programming. A total of 39 natural products underwent optimization via heuristic methodologies within the CODESSA software, resulting in the establishment of a linear model. Subsequent to this phase, a mere quintet of descriptors was curated for the generation of non-linear models through gene expression programming.

The squared correlation coefficients and s2 values derived from the heuristics stood at 0.5516 and 0.0195, respectively. Gene expression programming yielded squared correlation coefficients and mean square errors for the training set at 0.78 and 0.0085, respectively. For the test set, these values were determined to be 0.71 and 0.0121, respectively. The s2 of the heuristics for the training set was discerned to be 0.0085.

The analytic scrutiny of both algorithms underscores their commendable reliability in forecasting the efficacy of nascent compounds. A juxtaposition based on correlation coefficients elucidates that the GEP algorithm exhibits superior predictive prowess relative to the HM algorithm for novel synthetic compounds.