Update 2.1.1_host_prediction_I.md

_episodes/2.1.1_host_prediction_I.md

@@ -32,9 +32,26 @@ Listen to lecture by Varada and Malte (pay attention!!!)
 >    a) Two viral contigs match to the same host during the host prediction
 >    b) One viral contig matches 2 hosts: [s__Salinibacter ruber](https://gtdb.ecogenomic.org/tree?r=s__Salinibacter%20ruber) and [s__Longimonas halophila](https://gtdb.ecogenomic.org/tree?r=s__Longimonas%20halophila) 
 > {: .source}
-{: .discussion}
+{: .challenge}
 
-For the hands-on part - we will be using a tool called RaFAH.  [RaFAH paper](https://www.sciencedirect.com/science/article/pii/S2666389921001008?via%3Dihub).  
+For the hands-on part - we will be using a tool called RaFAH. Have a look into 
+the [RaFAH paper](https://www.sciencedirect.com/science/article/pii/S2666389921001008?via%3Dihub) 
+and focus on the introduction and Figures 1 and 2.  
 
-This tool uses random forest model machine learning models to predict hosts for phages. They train the ML model using manually curated classical host predictions (CRISPR sequences, tRNA and homology based matches). 
+This tool uses random forest model machine learning models to predict hosts for phages. They train 
+the ML model using manually curated classical host predictions (CRISPR sequences, tRNA and homology based matches). 
+
+
+> ## Questions
+> {:start='6'}
+> 6. Briefly discribe how RaFAH derives its predictions from a genome sequence.
+>
+> RaFAH produces a probability score for each host genus in its training set. In Figure 2,
+> you can find, how this score relates to precision and recall for the test dataset the authors
+> used.
+> 
+> {:start='7'}
+> 7. Decide, which probability score you would use as a cutoff. Explain your decision.
+> {: .source}
+{: .challenge}