ASAP Discovery x OpenADMET CompetitionTake part in the first prospective benchmark on Polaris.

Benchmark

polaris/pkis1-ret-wt-mut-r-1

Multitask classification benchmark for RET wild type, mutant V804L, and mutant Y791F.

Created on: December 08, 2023Train size: 279Test size: 85
Public
Multi Task
Regression

Participants

Tags

kinase
hit-discovery
selectivity

Leaderboard

Test set

Task

#
NameContributors
mean_absolute_error
mean_squared_error
r2
spearmanr
pearsonr
explained_var
References
1

pkis1-ret-wt-mut-r-1_DGLGraphTransformer_MPNNModel

20.518
709.711
-0.012
-0.239
-0.236
-0.000
No references provided
2

pkis1-ret-wt-mut-r-1_DGLGraphTransformer_AttentiveFPModel

20.840
733.827
-0.047
0.035
-0.000
-0.029
No references provided
3

pkis1-ret-wt-mut-r-1_DGLGraphTransformer_GCNModel

17.626
746.647
-0.065
0.491
0.554
0.222
No references provided
4

pkis1-ret-wt-mut-r-1_DGLGraphTransformer_PNAModel

25.118
833.102
-0.188
0.244
0.316
0.065
No references provided
5

pkis1-ret-wt-mut-r-1_avalon_FCModel

19.366
953.045
-0.359
-0.084
-0.116
-0.027
No references provided
6

pkis1-ret-wt-mut-r-1_DGLGraphTransformer_GATModel

20.562
1103.065
-0.573
0.099
0.074
0.003
No references provided

Details

README

Background

RET (Rearranged during Transfection) is a proto-oncogene that codes for a receptor tyrosine kinase. This means it produces a protein that plays a role in signaling pathways within cells, particularly related to cell growth and differentiation. When activated, RET helps regulate cell survival, proliferation, and differentiation. Mutations or alterations in the RET gene can lead to uncontrolled cell growth and potentially the development of cancer.

Benchmarking

RET wild type: In some cases, targeting both mutant and wild-type RET together can be more effective than targeting only one form as Combination Therapies. In certain cancer types, such as some subtypes of non-small cell lung cancer (NSCLC), the RET signaling pathway can interact with other oncogenic pathways, such as the EGFR (epidermal growth factor receptor) pathway. Targeting both pathways simultaneously might offer a synergistic effect and improve treatment outcomes.

RET-V804L: The V804L mutation causes a structural change in the RET protein, resulting in its continuous activation, even in the absence of ligand binding. The mutation leads to uncontrolled cell growth and division, contributing to oncogenesis. The V804L mutation in RET has been identified in various cancer types, particularly in thyroid cancers, including papillary thyroid carcinoma (PTC) and medullary thyroid carcinoma (MTC). It is often associated with aggressive tumor behavior and resistance to conventional therapies.

RET-Y791F: The Y791F mutation disrupts a crucial phosphorylation site within the RET protein. Consequently, the tyrosine at position 791 cannot be phosphorylated effectively or at all. This disruption interferes with the normal signaling pathways that rely on this particular phosphorylation event, leading to dysregulated downstream signaling, such as MAPK and PI3K. The Y791F mutation, by abrogating this phosphorylation site, alters these cellular responses and may influence tumor development and progression. The Y791F mutation is found in the RET protein, which is implicated in several types of cancer, particularly medullary thyroid carcinoma (MTC).

The goal of this benchmark is to select the best predictive model for

  • Selectivity towards the mutants
  • Optimization of the bioactivity % inhibition.
  • Discovery of potential hits in new chemical space.

Description of readout

Description of readout:

  • Readouts: RET, RET_(V804L_mutant), RET_(Y791F_mutant)
  • Bioassay readout: Percentage of inhibition.
  • Optimization objective: Higher value (higher %inhibition ).
  • Number of data points: train: 279 test: 85

Data resource:

Train/test split

Given the benchmarking goal, a scaffold-based splitting approach was applied to ensure training and test sets contain distinct chemical structures while maintaining the diversity of scaffolds.

Distribution of the train/test in the chemical space image

For more details of this benchmark -> notebook