FLIGHTED: Inferring Fitness Landscapes from Noisy High-Throughput Experimental Data (Part 2)

Data for FLIGHTED (Inferring Fitness Landscapes from Noisy High-Throughput Experimental Data). This data contains the TEV protease landscape and models trained on it. All other FLIGHTED data is in the other Zenodo repository (refer to the paper for details).

The data is arranged in the following folders:

TEV_Landscape: contains the TEV landscape (in flighted_fitnesses.csv) and splits thereof in Splits/. The main files for model training are flighted_fitnesses.csv and the files labeled one_vs_rest, two_vs_rest, and three_vs_rest. The files labeled three_vs_rest_control within Splits/ and the read count CSV files refer to further information about the read count in the landscape; see the Supplement for details. The dictionary files are the original raw data prior to processing with FLIGHTED.

TEV_Models: contains models trained on the TEV landscape under the various splits. Each model folder contains hyperparameters, training history, and predictions on the test set which can be used to evaluate model performance. Raw model parameters are not provided for fine-tuned models due to size; contact us if you want them. The control_run/ refers to the run described in the supplement on just read counts.