Avantika Lal

Pretrained model for bulk low-quality data

6

Hi @avantikalal, in cases where noisy samples have coverage >= that of the clean samples, should users _always_ forego training and use your pretrained model, nvidia:atac_bulk_lowqual_20m_20m? For example:  If...

Fix to include ends of chromosomes.

2

Currently get_intervals.py deletes intervals that extend beyond the chromosome length. This means that the last few kilobases of each chromosome are not covered in the intervals and no prediction is...

Include a set of config files with commonly used parameters, covering single-GPU and multi-GPU use. Give each config file a descriptive name and allow users to select a suitable config...

Add evaluation script for quality enhancement use case

1

Save array for TSS enrrichment plot FRIP equivalent (Fraction of Signal in Peaks?) - based on called peaks or peaks in clean data or both? Number of peaks Number of...

Could not construct partition: Cannot accept NaN weights.

1

Hi, using the following command, I get an error: ``` pos_patterns, neg_patterns = modiscolite.tfmodisco.TFMoDISco( hypothetical_contribs=attrs, one_hot=inputs, max_seqlets_per_metacluster=2000, sliding_window_size=20, flank_size=5, target_seqlet_fdr=0.05, n_leiden_runs=2, ) ``` The error message is below: ``` /opt/conda/lib/python3.10/site-packages/modiscolite/affinitymat.py:238:...

I get the error "UnboundLocalError: local variable 'n_cells' referenced before assignment" on trying to run the "fragments_to_sparse" function without providing a list of cells to include. I think it's caused...

hit scoring

2

Hi @jmschrei , We'd like to try hit scoring (shown in the section "Do hit scoring" of the modisco tutorial: https://github.com/kundajelab/tfmodisco/blob/master/examples/simulated_TAL_GATA_deeplearning/TF_MoDISco_TAL_GATA.ipynb) using outputs from modiscolite. Is this implemented, and if...