stochprofML: stochastic profiling using maximum likelihood estimation in R

BMC Bioinformatics

Table 2 Estimates of numbers of cells from the first population in the simulated 5-cell data described in Figs. 12 and 13a and in the main text

Estimator for # of cells in pop. 1		Measurement index						# of hits
Estimator for # of cells in pop. 1		1	2	3	4	5	6	# of hits
Estimated parameters	Mean	0.00	1.00	1.00	1.00	2.14	1.01	98
Estimated parameters	MLE (CI)	0 (0,0)	1 (1,1)	1 (1,1)	1 (1,1)	2 (2,3)	1 (1,1)	98 (100)
True parameters	Mean	0.00	1.00	1.00	1.00	2.39	1.02	97
True parameters	MLE (CI)	0 (0,0)	1 (1,1)	1 (1,1)	1 (1,1)	2 (2,3)	1 (1,1)	97 (100)
True # of cells from population 1		0	1	1	1	2	1

Columns: Estimation results for the first six measurements from the datasets and (last column) summary across all 100 samples. Rows: Estimation of cell numbers are based on conditional probabilities that use either the estimated model parameters (rows 1 and 2, corresponding to blue bars in Fig. 13a) or the true values (rows 3 and 4, orange bars). Within each of these two choices one can consider the mean number of cells from population 1 as determined by the conditional probabilities (rows 1 and 3) or the maximum likelihood estimator (MLE) that maximizes the conditional probabilities (rows 2 and 4, first value) including a 95% confidence interval that covers at least 95% of the conditional probability mass (rows 2 and 4, in parentheses). The last row shows the true pool composition. The last column shows for each estimator how many of the 100 cell numbers were inferred correctly (defined as follows: rounded mean is exact match; MLE is exact match; confidence interval (CI) includes correct number)

ISSN: 1471-2105