ss estimate refactor

[Ukyeon]

1. optimized the multiprocessing by switching from ThreadPool to a process pool that utilizes cores instead of threads. This change significantly improved the operating time:

   • Previous operating time: 5.573467457999996
   • After optimization time : 2.393449375000003

2. removed some duplicated code and simplified the existing code.

[Xiaojieqiu]

this is excellent, Ukyeon! Can you please show me some benchmark on this? Let us also confirm the resulting parameters are the same after your updates?

[Ukyeon]

The resulting parameters are the same after the updates..

est = ss_estimation() # After est2 = ss_estimation_prev() # Before

np.array_equal(est.parameters["gamma"], est2.parameters["gamma"]) True np.array_equal(est.aux_param["gamma_intercept"], est2.aux_param["gamma_intercept"]) True np.array_equal(est.aux_param["gamma_r2"], est2.aux_param["gamma_r2"]) True np.array_equal(est.aux_param["gamma_logLL"], est2.aux_param["gamma_logLL"]) True np.array_equal(est.aux_param["bs"], est2.aux_param["bs"]) True np.array_equal(est.aux_param["bf"], est2.aux_param["bf"]) True

// After adata.uns["dynamics"] {'filter_gene_mode': 'final', 't': None, 'group': None, 'X_data': None, 'X_fit_data': None, 'asspt_mRNA': 'ss', 'experiment_type': 'conventional', 'normalized': True, 'model': 'stochastic', 'est_method': 'gmm', 'has_splicing': True, 'has_labeling': False, 'splicing_labeling': False, 'has_protein': False, 'use_smoothed': True, 'NTR_vel': False, 'log_unnormalized': True, 'fraction_for_deg': False}

// Before adata.uns["dynamics"] {'filter_gene_mode': 'final', 't': None, 'group': None, 'X_data': None, 'X_fit_data': None, 'asspt_mRNA': 'ss', 'experiment_type': 'conventional', 'normalized': True, 'model': 'stochastic', 'est_method': 'gmm', 'has_splicing': True, 'has_labeling': False, 'splicing_labeling': False, 'has_protein': False, 'use_smoothed': True, 'NTR_vel': False, 'log_unnormalized': True, 'fraction_for_deg': False}