Skip to content

subtype

Subtype 

Subtype(
    generator: GeneratorAD,
    n_subtypes: int = 2,
    GPU: Union[bool, str] = True,
    random_state: Optional[int] = None,
)
Source code in src\stands\subtype.py 
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
def __init__(self, 
             generator: GeneratorAD, 
             n_subtypes: int = 2,
             GPU: Union[bool, str] = True, 
             random_state: Optional[int] = None,
        ):

    self.n_subtype = n_subtypes
    self.device = select_device(GPU)
    self.G = generator.to(self.device)
    self.C = Cluster(self.G, self.n_subtype).to(self.device)

    if random_state is not None:
        seed_everything(random_state)
    self.seed = random_state

    generator = None  # remove model

fit 

fit(data: Dict[str, Any])

Detect subtypes of samples

Source code in src\stands\subtype.py 
29
30
31
32
33
34
35
36
37
38
def fit(self, data: Dict[str, Any]):
    '''Detect subtypes of samples'''

    graph = data['graph'].to(self.device)

    self.G.eval()
    self.C.train()
    z, res_z = self.generate_z_res(graph)
    q = self.C.fit(z, res_z)
    return q

generate_z_res 

generate_z_res(graph: DGLGraph)

Generate reconstructed data

Source code in src\stands\subtype.py 
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
@torch.no_grad()
def generate_z_res(self, graph: dgl.DGLGraph):
    '''Generate reconstructed data'''
    self.G.eval()
    if self.G.extract.only_ST:
        z, fake_g = self.G.STforward(graph, graph.ndata['gene'])
        res_g = graph.ndata['gene'] - fake_g.detach()
        res_z = self.C.STforward(graph, res_g)

    else:
        z, fake_g, fake_p = self.G.Fullforward(
            graph, graph.ndata['gene'], graph.ndata['patch']
        )
        res_g = graph.ndata['gene'] - fake_g.detach()
        res_p = graph.ndata['patch'] - fake_p.detach()
        res_z = self.C.Fullforward(graph, res_g, res_p)

    return z, res_z