bkoep 91 885 
Offline
This is presumably the cytoplasmic part of ORF3a (UniProt P59632 for the very similar SARS-Cov-1 version). In line with this environment, no disulfides are expected by DiaNNA.
This thing is somewhat conserved in other betacov genomes; bat versions include A0A088DI21 and A0A1B3Q5V5. Below is an alignment of these guys; you can see the human-infecting SARSr-COVs have some extensions not found in the two bat versions.
>Pz1844 WKCRSKNPLLYDANYFLCWHTN CYDYCIPYNSVTSSI-VITSGDGTTSPISEHDYQIGGYTEKWESGVKDCVVLHSYFTSDY YQLYSTQLSTDTGVEHVTFFIYNKIVDEPEEHVQIHTIDGSSGVVNPVMEPIYDEPTTTCT SVP- >New|UniRef100_A0A088DI21 Uncharacterized protein n=1 Tax=Bat Hp-betacoronavirus/Zhejiang2013 TaxID=1541205 RepID=A0A088DI21_9BETC IRCKSLVPLCADDDCFVNYNAG GKTYCMPFDPNEPYLTLVVHQNGIT---------CGSYKLYGDVSIADRIYLVTLTKSVP YSLQNI---FDAELCTIAFYIADCAV------IEDHTTAGKTPRLELKSDPIYEVPCATI DVPL >New|UniRef100_A0A1B3Q5V5 NS3 protein n=1 Tax=Rousettus bat coronavirus TaxID=1892416 RepID=A0A1B3Q5V5_9BETC IRVHSMAPFVSTADNFAVLRTT CSRFVFPVESSKDNVVVLTTSRGVF---------CNGIHVEGPTALSDNASIVSLFSTTV LLLDRVEQGYDY---TVFVYISQQILRNSE--------SNPQGVVNPEFD---------- DVEL
This probably won't work, but just for the sake of fun I told Rosetta's GREMLIN to try and figure out what residues are changing together. This sort of correlation is usually indicative of close contact in the actual protein. http://openseq.org/sub.php?id=1591034373 The last time GREMLIN tried it was in 2013 without enough sequences to work with, and I don't really see that changing much: http://gremlin.bakerlab.org/pfam.php?id=PF11289&year=2013.
