The Unreasonable Effectiveness of Inverse Reinforcement Learning in Advancing Cancer Research
Authors: John Kalantari, Heidi Nelson, Nicholas Chia437-445
AAAI 2020 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | In testing PUR-IRL on real-world data from colorectal cancer (CRC) patients, we verify its ability to infer a series of mutational events, or an evolutionary trajectory, that broadly matches those arrived at by CRC experts through the curation of a variety of multi-omics and experimental data sources. Furthermore, we show that PUR-IRL can accomplish this with data taken from a mere tens of patients and that this outperforms frequency-based statistical approaches that are commonly used in cancer research. Our experimental results show that PUR-IRL can correctly identify the number of distinct experts, the reward function and optimal policy of each expert, and remain robust in classification under various data sampling conditions. Tested on Grid World, PUR-IRL achieved an F1-score of 0.9328 and 0.90331 under uniform and non-uniform sampling conditions, respectively. |
| Researcher Affiliation | Academia | John Kalantari,1,2 Heidi Nelson,1,3 Nicholas Chia1,2,4 1Microbiome Program, Center for Individualized Medicine, Mayo Clinic, Rochester, MN, USA 2Division of Surgical Research, Department of Surgery, Mayo Clinic, Rochester, MN, USA 3Division of Colon and Rectal Surgery, Department of Surgery, Mayo Clinic, Rochester, MN, USA 4Department of Physiology and Biomedical Engineering, Mayo Clinic, Rochester, MN, USA {kalantari.john, chia.nicholas}@mayo.edu |
| Pseudocode | Yes | Algorithm 1 PUR-IRL |
| Open Source Code | No | The paper does not provide a specific link or explicit statement about the general availability of the source code for the methodology described. |
| Open Datasets | Yes | Whole genome sequencing (WGS) was performed on samples from a previously described study (Hale et al. 2018a; 2018b) and An action corresponds to an event occurring at one of 1084 known driver genes of CRC aggregated from two public datasets (Bamford et al. 2004; Tomczak, Czerwi nska, and Wiznerowicz 2015). |
| Dataset Splits | No | The paper does not explicitly provide specific percentages, counts, or a clear methodology for training/validation/test dataset splits. It mentions 'uniform and non-uniform sampling conditions' and '3 expert demonstrations generated per expert' for Grid World, but no explicit splits. |
| Hardware Specification | No | The paper mentions 'high-performance computing clusters' in the acknowledgments but does not provide specific details such as GPU models, CPU types, or memory specifications used for experiments. |
| Software Dependencies | No | The paper lists software tools like 'Mu Tect2', 'Filter Mutect Calls from the Genome Analysis Toolkit (GATK)', 'Titan', 'SNPeff', and 'Phylo WGS' but does not specify their version numbers. |
| Experiment Setup | Yes | The PUR-IRL model was run with 6 pop-up updates between every 100 CRP iterations. We evaluate the IRL model under different discount hyperparameter values {0.0, 0.3, 0.7, 1.0} |