Automatic Cell Type Annotation Using Marker Genes for Single-Cell RNA Sequencing Data

. 2022 Oct 21;12(10):1539. doi: 10.3390/biom12101539

Algorithm 1 ACAM: Automatic Cell type Annotation Method.

Input:
The pre-processed data matrix X, marker gene set $G$
1:
Initialize $t h r e s h = 10$ , $k = 1$
2:
$C_{1} \leftarrow S C 3 (X)$ ; $C_{2} \leftarrow C I D R (X)$ ; $C_{3} \leftarrow S e u r a t (X)$ ;

$C_{4} \leftarrow t$ - $S N E + k$ - $m e a n s (X)$ ; $C_{5} \leftarrow S I M L R (X)$
3:
$R_{i, j} \leftarrow A R I (C_{i}, C_{j}), i = 1, \dots, 5$
4:
Remove the method of $arg {min}_{i} v a r (R_{i, \cdot})$
5:
$A_{i}$ : $A_{i} (u, v) \leftarrow \{\begin{matrix} 1, & u, v from the same cluster, \\ 0, & otherwise . \end{matrix}$
6:
$\tilde{A} \leftarrow \sum_{i = 1}^{4} A_{i}$
7:
$A^{c o n}$ : $A^{c o n} (u, v) \leftarrow {\begin{matrix} 1, & \tilde{A} (u, v) = 4, \\ 0, & else . \end{matrix}$
8:
$P = {P_{1}, P_{2}, \dots, P_{C}} \leftarrow L o u v a i n (A^{c o n}, t h r e s h)$
9:
for c = 1,…,C do
11:
${\tilde{P}}_{c} \leftarrow O v e r s a m p l e (P_{c})$
11:
Select m with $m e a n (X [m, {\tilde{P}}_{c}]) > m e a n (X [m, U - P_{c}])$
12:
$w_{m} \leftarrow X G B o o s t ({\tilde{P}}_{c}, U - P_{c}), m \in G_{t}, t = 1, \dots T$
13:
$S c o r e_{t} = \sum_{m \in G_{t}} w_{m}, t = 1, \dots, T$
14:
$y_{u} \leftarrow cell type t_{0}$ : $t_{0} = arg {max}_{t} (S c o r e_{t}), u \in P_{c}$
15:
end for
16:
$y_{u} \leftarrow k N N (X [, P], k), u \in U - P$
Output:
Cell labels y