diff --git a/doc/iss/generate.rst b/doc/iss/generate.rst
index 376c0c6..c6f6c50 100644
--- a/doc/iss/generate.rst
+++ b/doc/iss/generate.rst
@@ -135,13 +135,28 @@ Coverage distribution
 
 In the context of InSilicoSeq, the `abundance` is the proportion of reads in a sample, which since it does not acount for the length of the genome, does not necessarily reflect the number of organisms present in a sample.
 
-The `coverage`and `coverage_file` options allow for simulating reads according to a coverage distribution instead of abundance.
+The ``coverage`` and ``coverage_file`` options allow for simulating reads according to a coverage distribution instead of abundance.
 
 .. code-block:: bash
 
     iss generate --ncbi bacteria -U 50 --coverage lognormal -n 25M \
         --model novaseq --output reads
 
+The ``coverage_file`` option works similarly to the ``abundance_file`` option.
+For two genomes A and B:
+
+.. code-block:: bash
+
+    iss generate --genomes genomes.fasta --coverage_file coverage.txt \
+        --model novaseq --output reads
+
+with, for a coverage of 20x for genome_A and 100x for genome_B, the coverage file `coverage.txt` will be:
+
+.. code-block:: bash
+
+    genome_A    20
+    genome_B    100
+
 GC bias
 -------
 
diff --git a/iss/app.py b/iss/app.py
index 0dce452..24b7d8a 100644
--- a/iss/app.py
+++ b/iss/app.py
@@ -229,7 +229,8 @@ def generate_reads(args):
             f = open(genome_file, 'r')  # re-opens the file
             with f:
                 fasta_file = SeqIO.parse(f, 'fasta')
-
+                total_reads_generated = 0
+                total_reads_generated_unrouned = 0
                 for record in fasta_file:
                     # generate reads for records
                     try:
@@ -252,9 +253,22 @@ def generate_reads(args):
                                 err_mod.read_length,
                                 genome_size
                             )
-                        n_pairs = int(round(
-                            (coverage *
-                                len(record.seq)) / err_mod.read_length) / 2)
+                        n_pairs_unrounded = (
+                            (coverage * len(record.seq)) /
+                            err_mod.read_length) / 2
+                        n_pairs = round(n_pairs_unrounded)
+
+                        # check that the rounding does not cause to drop
+                        # read pairs
+                        total_reads_generated_unrouned += n_pairs_unrounded
+                        total_reads_generated += n_pairs
+                        if round(total_reads_generated_unrouned) > \
+                                total_reads_generated:
+                            logger.debug(
+                                "Adding a pair to correct rounding error")
+                            n_pairs += 1
+                            total_reads_generated += 1
+
                         # skip record if n_reads == 0
                         if n_pairs == 0:
                             continue
diff --git a/iss/version.py b/iss/version.py
index 51ed7c4..c3b3841 100644
--- a/iss/version.py
+++ b/iss/version.py
@@ -1 +1 @@
-__version__ = '1.5.1'
+__version__ = '1.5.2'