eternagame / eternafold Goto Github PK

Expected Behavior

Probability of pairing $p_i$, calculated as a sum of all pairing probabilities $p_{ij}$, should not exceed unity, $p_i = \sum_j p_{ij} \leq 1$.

Current Behavior

For longer sequences, probability of pairing for certain nucleotides exceeds 1, the difference being significantly greater than one would expect from numerical instabilities.

Reproduction Steps

Fold the following MS2 sequence (MS2_short.txt) with default parameters:

./src/contrafold predict MS2_short.txt --params parameters/EternaFoldParams.v1 --posteriors 0.00001 MS2_bps.txt

Looking at the obtained pair probability matrix (MS2_bps.txt), one observes the following two lines:

571 C 613:0.00438902 616:0.996947
572 C 612:0.00449777 615:0.996294

The total pairing probability for nucleotides 571 and 572 therefore exceeds 1: $p_{571} = 1.00134$, $p_{572}= 1.000792$.

Environment

No response

Other information

No response