simon-anders

library( irlba )

m <- 10000 # nbr of features (rows)
n <- 5000  # nbr of cells (colums)
r <- 5  # nbr of latent components

## The true latent values

# True importance of latent factors
true_importance <- c( 1, .8, .4, .2, .1 )

simon-anders

## Beispiel für Plot-Größe

Tidyverse laden:

```{r}
library( tidyverse )
```

Wir benutzen `mtcars`, eine Standard-Beispiel-Tabelle von R mit technischen Daten 
für (recht alte) Autos:

simon-anders

import gzip, random

# Load FASTQ file for Chromosome 10 from GRCm38
with gzip.open("data/Mus_musculus.GRCm38.dna.chromosome.10.fa.gz") as f:
    firstline = f.readline()
    assert firstline.startswith(b'>')
    chrom_seq = b"".join(l.rstrip() for l in f)

# This here is the file from papagei:mnt/raid/scnmt_data/CpG_filtered
cpg = scipy.sparse.load_npz( "data/CpG_10.npz" ).tocoo()

simon-anders

# Make example data
n <- 15
x <- runif( n, 0, 10 )
y <- sin(x) + rnorm( n, sd=.1 )

# Make grid to plot smooth curves
xg <- seq( 0, 10, length.out=1000 )

# Determine knot positions:

simon-anders

library( rstan )
options(mc.cores=6)

n <- 1000
s <- round( 10^rnorm( n, 3, .5 ) )
fracs <- 10^ifelse( runif(n)<.7, rnorm( n, -3.3, .4 ), rnorm( n, -1.7, .2 ) )
k <- rpois( n, fracs*s )

model <- stan_model( model_code="
data {

simon-anders

---
format:
  html:
    embed-resources: true
---    

# Density estimation for single-cell gene expression with MCMC

Simon Anders, 2024-10-23