The mistake was you playing for a win in a drawn position.
A common error...trust me on this. 🙁
Here you played 43... Ka5 trying to push home your passed pawn.
Now the Black King cannot get into the White's e-pawn 'square' so then
a timed d4-d5 will create a passed e-pawn which the Black King after exd5
and e6 will not be able to stop.
So it looks like 43....Kc6 giving up all hope of winning stops all the d4-d5 tricks.
After you played 43....Ka5 it played 44.g6.
Your hackles your have risen here. Computers do not play moves like this unless
it's to stop some deep human unfathomable checkmate in 20 moves or it has done
some cold calculation. This looks like the latter.
So having determined you probably do not have a checkmate in 20 moves you
have to stop and figure out why it played 44.g6. These things do not blunder pawns.
It is obviously planning d4-d5 and a pawn on g6 of any colour stops or slows down
Be2 and Bh5 moves which allows the Bishop to cover the queening square e8.
You played 45...Ba2 and it now has an easy win with 46.Kb2 swapping
pieces and playing d4-d5. The King can run back instead of taking of d5
but after dxe6. The Black King has to take to long way round to get to e7
due to the double e-pawns controlling d6 and d7.
Meanwhile the White King picks up the b-pawn and plays Kc5.
45...Be2 had to be tried. In answer to 46.d5 exd5 47.e6 you may
have b4+ and Bb5. I cannot see a 100% White win but fear the worst,
I've no idea how deep it went when playing g6.
44.Ne3 without g6 still wins if the Bishop goes to a2 but it
would not have been expecting that. It's a blunder.