add more notes about scaled

notebooks/20220913-fraction-mtx-after-subtracting-mgx.ipynb

@@ -11,9 +11,15 @@
     "\n",
     "These values were estimated by using FracMinHash sketching. \n",
     "[FracMinhash sketches](https://www.biorxiv.org/content/10.1101/2022.01.11.475838v2.abstract) represent a fraction of the k-mers in the original sequence.\n",
+    "\n",
     "K-mers are words of length _k_ in nucleotide sequences.\n",
     "Here we used k-mer sizes 21, 31, and 51 because these can be used to estimate genus, species, and strain-level similarity respectively between sets of sequences.\n",
-    "We also downsampled to 1/200th of all k-mers to facilitate faster comparisons.\n",
+    "\n",
+    "We also downsampled to 1/200th (0.5%) of all k-mers to facilitate faster comparisons (`scaled = 200`).\n",
+    "Importantly, the same fraction of k-mers is retained across samples if the same sequences are contained within those samples.\n",
+    "Within the sourmash ecosystem, scaled values of 1000 and 2000 are the most standard because these values are a nice compromise for speed of comparisons while retaining enough k-mers to accurately represent microbial genomes.\n",
+    "Because viral genomes can be very small, we reduced the scaled value to 200 to retain more k-mers with the idea that even small viral genomes would still have a few representative k-mers within the sketch.\n",
+    "See [this issue](https://github.com/sourmash-bio/sourmash/issues/1340) for more details about how the scaled value and the k-mer size impact the number of k-mers retained for viral genomes of different sizes.\n",
     "\n",
     "After we sketched both the metatranscriptome and the metagenome, k-mers that occurred in the metagenome were removed from its paired metatranscriptome.\n",
     "\n",