Added option (--equal-haps) to enforce that haplotype estimates are equivalent for transcripts in the same cluster when using the haplotype-transcript inference model. While this seem to overall improve results it is not the default mode yet as it can potentially result in worse performance for transcripts that only share little sequence.
Removed the path overlap-based clustering as default. Only paths that share reads with positive probability are now clustered into the same group. This generally results in faster downstream inference as the clusters becomes smaller. The old path overlap-based clustering can be re-activated using --path-node-cluster.
Changes to existing options and outputs:
Changed ClusterRelativeExpression and ReadCount in the abundance output so that it now calculates the mean across all haplotype samples. This also corresponds to multiplying the old values with the HaplotypeProbability.
Bug fixes:
Fixed bug resulting in incorrect normalization of read path probabilities (only had a minor influence on results).
Improvements:
--equal-haps
) to enforce that haplotype estimates are equivalent for transcripts in the same cluster when using the haplotype-transcript inference model. While this seem to overall improve results it is not the default mode yet as it can potentially result in worse performance for transcripts that only share little sequence.--path-node-cluster
.Changes to existing options and outputs:
Bug fixes: