labomics / midas Goto Github PK

as the title

Hi, I have two questions:

1. I noticed that using --act predict_all_latent_bc can output x_bc folder which contains the imputed missing modality features. But It seems MIDAS has a mode named translate which can also impute missing modality features. If I have one batch that's only measured with RNA modality and I wanna impute its ATAC features. Which --act should I use? Can i directly use the x_bc output?
2. If I wanna evaluate the imputed features with ground truth, can i use the outputs in x_bc folder?

Thanks.

Hi, I have been trying to reproduce the results of midas on Dogma dataset. I have completed the preprocessing as intructed in the docs. But when I try to run the method, I get the following error:

$ CUDA_VISIBLE_DEVICES=0 python3 run.py --exp e0 --task dogma_single_full &
$ Task: dogma_single_full
Experiment: e0
Model: default
Input feature numbers: {'atac': 27489, 'rna': 4041, 'adt': 208}
Total mini-batch size: 256, GPU number: 1, GPU mini-batch size: 256
Parameter number: 17.756 M
[0] {0}
Traceback (most recent call last):
File "run.py", line 707, in
main()
File "run.py", line 107, in main
train()
File "run.py", line 327, in train
train_data_loader_cat = get_dataloader_cat("train", train_ratio=None)
File "run.py", line 359, in get_dataloader_cat
datasets.append(MultimodalDataset(o.task, o.reference, o.data_dir, subset, split, train_ratio=train_ratio))
File "/home/vineet/UGP/midas/modules/datasets.py", line 45, in init
assert cell_nums[0] > 0 and len(set(cell_nums)) == 1,
AssertionError: Inconsistent cell numbers!

Can you explain, why I might be getting this error?

Thank you for your fantastic design of the MIDAS software for mosaic integration and imputation of single-cell multi-modal data.

Our team is conducting a registered study on multi-modal integration; however, we are unsure whether your MIDAS software has been released yet.

We hope for your quick reply, as it would greatly aid our study.

Thanks,

Line 225, run.py link

o.dims_h[m] = dim if m != "atac" else o.dims_enc_chr[-1] * 22

should be

o.dims_h[m] = dim if m != "atac" else o.dims_enc_chr[-1] * len(o.dims_chr)

I want to reproduce the plot shown in Fig. 2 (h.i.j), but there are no more details here. Can you provide the code to draw this picture? thank you very much

Regards,
Jiawei

Hi, I met a problem when applying MIDAS to my own dataset: one batch measured with RNA+histone modafication (H3K27me3) modalities and another batch measured with RNA+ATAC modalities. I defined three modalities as input: rna, atac, and H3K27me3. Following the steps for processing atac data, I split H3K27me3 into chunks by chromosome id. But the code cannot run correctly (I can correctly reproduce the RNA+ATAC integration experiment). It seems that MIDAS only recognizes chunk inputs for 'atac' modality, although 'H3K27me3' mod is very similar 'atac' mod. On the other hand, taking the H3K27me3 feature (>100,000) as a whole will significantly increase the computational complexity.

How can I solvle this problem?

Dear Prof. Ying

  Recently, I had the pleasure of reading your research work titled 'Mosaic integration and knowledge transfer of single-cell multimodal data with MIDAS'. The novelty and effectiveness of this work deeply attracted me. I've started trying to use MIDAS to process my mosaic data. However, after reading your tutorial, I still don't understand how to use MIDAS to analyze my data. My data are in common file formats; for example, RNA and ATAC data are in three files at 10x: barcodes.tsv, features.tsv, and matrix.mtx. Protein abundance data is in a CSV file. Can you provide a tutorial that matches this type of file format? Or perhaps a tutorial compatible with Seurat or Scanpy?"