Update README.md

Author: jmschrei

README.md

@@ -5,7 +5,7 @@
 [[docs](https://tangermeme.readthedocs.io/en/latest/index.html)][[tutorials](https://github.com/jmschrei/tangermeme/tree/main/docs/tutorials)][[vignettes](https://github.com/jmschrei/tangermeme/tree/main/docs/vignettes)]
 
 > [!NOTE] 
-> tangermeme is under active development. The API has largely been decided on, but may change slightly across versions until the first major release.
+> If you use tangermeme in your work, please consider citing the [preprint](https://www.biorxiv.org/content/10.1101/2025.08.08.669296v2). Citations allow me to continue developing software like this for the community. 
 
 Training sequence-based machine learning models has become widespread when studying genomics. But, what have these models learned, and what do we even do with them after training? `tangermeme` aims to provide robust and easy-to-use tools for the what-to-do-after-training question. `tangermeme` implements many atomic sequence operations such as adding a motif to a sequence or shuffling it out, efficient tools for applying predictive models to these sequences, methods for dissecting what these predictive models have learned, and tools for designing new sequences using these models. `tangermeme` aims to be assumption free: models can be multi-input or multi-output, functions do not assume a distance and instead return the raw predictions, and when loss functions are necessary they can be supplied by the user. Although we will provide best practices for how to use these functions, our hope is that being assumption-free makes adaptation of tangermeme into your settings as frictionless as possible. All functions are unit-tested and implemented with both compute- and memory-efficient in mind. Finally, although the library was built with operations on DNA sequences in mind, all functions are extensible to any alphabet.