Different researchers have proposed different methods on how to process and interpret scRNA-seq data. Unfortunately, these different data processing pipelines lead to different downstream conclusions, and there is no consensus in the community on what is the correct approach.
Zero-imputation methods are widely applied to address non-biological zeros in scRNA-seq data. However, these methods can introduce artificial signals, skewing the results of downstream analysis to match initial assumptions rather than emulate the underlying biological processes. Several popular zero imputation techniques provide significantly varied results on the downstream network inference tasks over the same real-world scRNA datasets. Check out our initial investigation results in this paper presented at ICML 2024 ML4LMS workshop. Interested in the details, here is our code.