You are right about R14.
About R21, R24... and R22, R23, R25... they determine the gain of the non-inverting amplifier G2.
It is the same for G0 (at the left).
Try to analyze the amplifier G0 and find its gain (Vnoisy/Vin) after you let Vn=0 (Vn connected to ground so that R05 and R04 become parallel).
I hope you learnt how to find the gain of a non-inverting amplifier using opamp... right?
You will find out that if Vin is gounded, the gain (Vnoisy/Vn of G0) will be the same as the previous one.
Vnoisy = Vin*50 + Vn*50
Vnoisy = 50*(Vin + Vn)
G2 and G0 are similar, but Vin is now the output of G1.
Concerning what we see at IN- of G1:
In theory, the voltage at IN- is equal to the voltage of IN+ which is connected to ground. So it should be 0V but actually we notice that it is not. Why?
Vclean sends a current Vclean/R13 to the node IN-. A part of it goes to ground (through the source Vin) via R11 and the rest of it continues through R12 which creates a rather large voltage at G1 output. If we divide the noise voltage at G1 output by the noise voltage at IN- we get the actual G1 gain alone (called open loop gain of the opamp) but for the frequency of Vn (which is 4 KHz here). This gain is not infinity or too large as we assume in our approximate calculations. Typically, when the frequency increases, the gain decreases. You can test this by changing the 4 KHz to 8 KHz.
Note: To ease the reading on the traces, you can let Vin=0 (or replace its label at R11 with the ground symbol) since the test now is about Vn only.
Kerim