修改下面这段话让表达更清晰Based on the above observations we proposed a general SV caller method-SVcnn which can accurately call DELs INSs and INVs The reason why we didnt add the function of detecting DUPs to SVc

Based on our observations, we have developed a general SV caller method called SVcnn. This method accurately detects DELs, INSs, and INVs. We have not included the function of detecting DUPs in SVcnn because we consider DUPs as a special case of INSs. To identify DUPs, we can compare INS sequences with the reference and if an INS can find a similar sequence in the reference, it is considered a DUP.

SVcnn is a convolutional neural network (CNN) based method consisting of three parts. The first part identifies candidate SV regions from the bam file. The second part converts the candidate SV regions into images and builds the letnet model. The third part filters false SVs through the letnet model and outputs the final SVs.

We tested SVcnn and other callers on three read datasets (CHM13, HG002, HG00733) and found that SVcnn outperforms current methods with an improved F1-score of 2-8% when the read depth is greater than 5x. Furthermore, SVcnn can identify more multi-allelic SVs with fewer false SVs.