How to use LDA when the number of samples is smaller than the number of features?

[Five296]

Hello, I currently have a confusion. When conducting differential analysis of bacteria, I often face the challenge of having a limited number of samples (only around twenty) and an excessive number of annotated species, making it difficult to decide which method to use. In my understanding, LDA may require a large number of samples, greater than the number of annotated species, for optimal use. If my sample size is small, can tools like yours still be used?