Hello,
the SCTransform looks very good. I just check the method section of your GB paper. I am a little confused about the step 2. Would you like to explain a bit about it?
In the second step, we exploit the relationship of model parameter values and gene mean to learn global trends in the data. We capture these trends using a kernel regression estimate (ksmooth function in R). We use a normal kernel and first select a kernel bandwidth using the R function bw.SJ. We multiply this by a bandwidth adjustment factor (BAF, default value of 3, sensitivity analysis shown in Additional file 2: Fig. S4). We perform independent regularizations for all parameters (Fig. 2).
Hello, the SCTransform looks very good. I just check the method section of your GB paper. I am a little confused about the step 2. Would you like to explain a bit about it?
In the second step, we exploit the relationship of model parameter values and gene mean to learn global trends in the data. We capture these trends using a kernel regression estimate (ksmooth function in R). We use a normal kernel and first select a kernel bandwidth using the R function bw.SJ. We multiply this by a bandwidth adjustment factor (BAF, default value of 3, sensitivity analysis shown in Additional file 2: Fig. S4). We perform independent regularizations for all parameters (Fig. 2).
Best,