Update README.md

fix zenodo link

Author: kathyxchen

README.md

@@ -70,7 +70,7 @@ See `example_slurm_scripts/1_example_seqpred.slurm_gpu.sh` and `example_slurm_sc
 
 Sequence class scores can be obtained from Sei chromatin profile predictions. There are 2 types of scores that can be computed:
 
-- Raw sequence class scores: For sequences only. Raw sequence class scores are projection scores of chromatin profile predictions projected on the unit-length vectors representing each sequence class. This is an intermediate score originally developed for variant score prediction and is made available for use for developing downstream analyses or applications, such as using them as a sequence representation. **Note** our manuscript uses the Louvain community clustering, whole-genome sequence class annotation of the human genome whenever we apply sequence classes to reference genome sequences, and we encourage the use of these annotations over the raw sequence class scores when possible. Sequence class annotations for hg38 and hg19 (lifted over from hg38) are available for download from [this Zenodo record](10.5281/zenodo.7113989). 
+- Raw sequence class scores: For sequences only. Raw sequence class scores are projection scores of chromatin profile predictions projected on the unit-length vectors representing each sequence class. This is an intermediate score originally developed for variant score prediction and is made available for use for developing downstream analyses or applications, such as using them as a sequence representation. **Note** our manuscript uses the Louvain community clustering, whole-genome sequence class annotation of the human genome whenever we apply sequence classes to reference genome sequences, and we encourage the use of these annotations over the raw sequence class scores when possible. Sequence class annotations for hg38 and hg19 (lifted over from hg38) are available for download from [this Zenodo record](https://doi.org/10.5281/zenodo.7113988). 
 - Sequence class variant effect score (nucleosome-occupancy-adjusted): For variants only. Computed as alt - ref of the raw sequence class scores **adjusted for nucleosome occupancy, i.e. histone normalized**. To better represent predicted variant effects on histone marks, it is necessary to normalize for nucleosome occupancy (for example, a LoF mutation near the TSS can decrease H3K4me3 modification level while increasing nucleosome occupancy, resulting in an overall increase in observed H3K4me3 quantity). Therefore, for variant effect computation, we used the sum of all histone profile predictions as an approximation to nucleosome occupancy and adjusted all histone mark predictions to remove the impact of nucleosome occupancy change (nonhistone mark predictions are unchanged). See manuscript methods for more detail.
 
 #### Sequence prediction