Designed and implemented by Chris Talbot, based on gaia from Grundler et al., pre-print here, GitHub here.
fastgaia is a Python package implementation of gaia that can run up to 30x faster for large tree sequences. fastgaia offers a command-line interface and a comprehensive Python function for inferring the geographic locations of ancestors given a tree sequence. After installing, use fastgaia --help from terminal or fastgaia.infer_locations(...) in your Python console for more details.
Note: In simulation tests on a 20x20 plane, fastgaia infers coordinates of ancestors within 0.16 units of their correct location across all timescales - this is only 0.01 units worse than gaia. fastgaia is particularly accurate for very recent ancestors, inferring locations within 0.001 units of the locations inferred using gaia. However, it should still be noted that fastgaia provides improved efficiency at the expense of accuracy. While inferences for more nodes with more recent samples as descendants will be close in accuracy to inferences from gaia, inferences in the deep past will be increasingly inaccurate compared to gaia without sufficient temporal sampling. Accuracy of discrete state inference remains untested.
- Continuous Inference: Infers continuous locations of nodes based on provided sample locations or tree sequence individual location data.
- Discrete Inference: Infers discrete states of nodes using a transition cost matrix.
- Flexible Output: Outputs results in CSV format for easy analysis.
You can install fastgaia via pip:
pip install fastgaia