dedup_dots

def dedup(dots, hiccups_filter=True):
    newdots = []
    ress = list(sorted(set(dots['res'])))
    for chrom in sorted(set(dots['chrom1'])):
        chromdots = dots[dots['chrom1']==chrom].sort_values(['start1', 'start2']).reset_index(drop=True)
        for res in ress:
            chromdots['Supported_%s' % res] = (chromdots['res']==res)
        tree = spatial.cKDTree(chromdots[['start1', 'start2']]) #Not sure what's the best coordinate to use, of the strongest pixel, centroid, or the center of the dot?
        drop = []
        for i, j in tree.query_pairs(r=50000):
            ires = chromdots.at[i, 'res']
            jres = chromdots.at[j, 'res']
            chromdots.at[j, 'Supported_%s' % ires] = True
            chromdots.at[i, 'Supported_%s' % jres] = True
            if ress[-1] in (ires, jres) or abs(chromdots.at[j, 'start1']-chromdots.at[i, 'start1'])<=20000:
                if ires>jres:
                    drop.append(i)
                else:
                    drop.append(j)
        newdots.append(chromdots.drop(drop))
    deduped = pd.concat(newdots).sort_values(['chrom1', 'start1', 'start2'])
    if hiccups_filter:
        l = len(deduped)
        deduped = deduped[~((deduped['start2']-deduped['start1']>100000) & (~np.any(deduped[['Supported_%s'%res for res in ress[1:]]], axis=1)))]
        print(l-len(deduped), 'loops filtered out as unreliable %s resolution calls' % ress[0])
    return deduped