How to prepare the Genes_nmf_w_basis_example.RDS for generating MPs?

[Lualululu]

Input:

Genes_nmf_w_basis is a list in which each entry contains NMF gene-scores of a single sample. In our study we ran NMF using ranks 4-9 on the top 7000 genes in each sample. Hence each entry in Genes_nmf_w_basis is a matrix with 7000 rows (genes) X 39 columns (NMF programs)

Generate_Meta_Programs.R generates MPs from NMFs programs that were calculated for each sample using different ranks. The NMF programs were calculated per sample using the ‘NMF’ R package:
NMFs_per_sample = nmf(x = Expression_matrix, rank = 4:9, method="snmf/r", nrun = 10)
NMF programs are listed in Genes_nmf_w_basis, where each entry contains NMF gene-scores of a single sample.

1. "Using ranks 4-9 on the top 7000 genes in each sample"
   How to choose the top 7000 genes? Does this mean that the top 7000 for each rank are the same?

2. “NMF gene-scores”
   How to calculate the NMF gene-scores?

3.How to correlate cells with their corresponding MP? Do different MP correspond to different cell subtypes?

[alievakrash]

Hi,
I am also trying to link the NMFs to the MPs, I go the Table 2 from the paper, but is there any additional data to check which of the Robust NMFs were assigned to which MP?
Thanks!

[lisch7]

Hi,
Do you know the reasoning behind generating 39 programs for each tumor? What is the rationale behind the number 39?
Thanks!

[Phoenix12580]

Hi, 嗨， Do you know the reasoning behind generating 39 programs for each tumor? What is the rationale behind the number 39?您知道为什么为每个肿瘤生成 39 个程序背后的原因吗？39 这个数字的依据是什么？ Thanks! 谢谢！

rank = 4:9 4+6+7+8+9就是39了

[Aditya-Gautam20]

@Lualululu
In paper "Hallmarks of transcriptional intratumour heterogeneity across a thousand tumours",
Look into methods -> gene filtering

They take 7000 highest mean expression genes, same to be used for each rank.